Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,147 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+Analysis_code/2.eda_preproccesing.ipynb filter=lfs diff=lfs merge=lfs -text
+data/ASOS/2018.csv filter=lfs diff=lfs merge=lfs -text
+data/ASOS/2019.csv filter=lfs diff=lfs merge=lfs -text
+data/ASOS/2020.csv filter=lfs diff=lfs merge=lfs -text
+data/ASOS/2021.csv filter=lfs diff=lfs merge=lfs -text
+data/ASOS/2022.csv filter=lfs diff=lfs merge=lfs -text
+data/ASOS/2023.csv filter=lfs diff=lfs merge=lfs -text
+data/ASOS/asos.feather filter=lfs diff=lfs merge=lfs -text
+data/ASOS/asos_dup.feather filter=lfs diff=lfs merge=lfs -text
+data/ASOS/asos_fill_위경.feather filter=lfs diff=lfs merge=lfs -text
+data/ASOS/asos_merge.feather filter=lfs diff=lfs merge=lfs -text
+data/asos_air_merge.feather filter=lfs diff=lfs merge=lfs -text
+data/data_for_demo/df_busan.feather filter=lfs diff=lfs merge=lfs -text
+data/data_for_demo/df_daegu.feather filter=lfs diff=lfs merge=lfs -text
+data/data_for_demo/df_daejeon.feather filter=lfs diff=lfs merge=lfs -text
+data/data_for_demo/df_gwangju.feather filter=lfs diff=lfs merge=lfs -text
+data/data_for_demo/df_incheon.feather filter=lfs diff=lfs merge=lfs -text
+data/data_for_demo/df_seoul.feather filter=lfs diff=lfs merge=lfs -text
+data/data_for_modeling/df_busan.feather filter=lfs diff=lfs merge=lfs -text
+data/data_for_modeling/df_daegu.feather filter=lfs diff=lfs merge=lfs -text
+data/data_for_modeling/df_daejeon.feather filter=lfs diff=lfs merge=lfs -text
+data/data_for_modeling/df_gwangju.feather filter=lfs diff=lfs merge=lfs -text
+data/data_for_modeling/df_incheon.feather filter=lfs diff=lfs merge=lfs -text
+data/data_for_modeling/df_seoul.feather filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_1_busan.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_1_daegu.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_1_daejeon.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_1_gwangju.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_1_incheon.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_1_seoul.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_2_busan.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_2_daegu.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_2_daejeon.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_2_gwangju.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_2_incheon.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_2_seoul.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_3_busan.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_3_daegu.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_3_daejeon.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_3_gwangju.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_3_incheon.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan10000/ctgan10000_3_seoul.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_1_busan.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_1_daegu.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_1_daejeon.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_1_gwangju.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_1_incheon.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_1_seoul.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_2_busan.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_2_daegu.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_2_daejeon.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_2_gwangju.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_2_incheon.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_2_seoul.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_3_busan.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_3_daegu.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_3_daejeon.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_3_gwangju.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_3_incheon.csv filter=lfs diff=lfs merge=lfs -text
+data/data_oversampled/ctgan20000/ctgan20000_3_seoul.csv filter=lfs diff=lfs merge=lfs -text
+data/local_selected.feather filter=lfs diff=lfs merge=lfs -text
+data/대기질/2017_12.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2018/2018_1.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2018/2018_10.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2018/2018_11.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2018/2018_12.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2018/2018_2.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2018/2018_3.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2018/2018_4.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2018/2018_5.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2018/2018_6.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2018/2018_7.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2018/2018_8.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2018/2018_9.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2018.feather filter=lfs diff=lfs merge=lfs -text
+data/대기질/2019/2019_01.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2019/2019_02.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2019/2019_03.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2019/2019_04.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2019/2019_05.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2019/2019_06.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2019/2019_07.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2019/2019_08.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2019/2019_09.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2019/2019_10.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2019/2019_11.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2019/2019_12.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2019.feather filter=lfs diff=lfs merge=lfs -text
+data/대기질/2020/2020_1.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2020/2020_10.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2020/2020_11.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2020/2020_12.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2020/2020_2.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2020/2020_3.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2020/2020_4.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2020/2020_5.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2020/2020_6.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2020/2020_7.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2020/2020_8.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2020/2020_9.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2020.feather filter=lfs diff=lfs merge=lfs -text
+data/대기질/2021/2021_1.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2021/2021_10.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2021/2021_11.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2021/2021_12.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2021/2021_2.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2021/2021_3.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2021/2021_4.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2021/2021_5.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2021/2021_6.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2021/2021_7.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2021/2021_8.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2021/2021_9.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2021.feather filter=lfs diff=lfs merge=lfs -text
+data/대기질/2022/2022_1.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2022/2022_10.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2022/2022_11.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2022/2022_12.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2022/2022_2.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2022/2022_3.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2022/2022_4.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2022/2022_5.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2022/2022_6.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2022/2022_7.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2022/2022_8.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2022/2022_9.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2022.feather filter=lfs diff=lfs merge=lfs -text
+data/대기질/2023/2023_1.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2023/2023_10.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2023/2023_11.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2023/2023_12.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2023/2023_2.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2023/2023_3.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2023/2023_4.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2023/2023_5.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2023/2023_6.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2023/2023_7.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2023/2023_8.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2023/2023_9.xlsx filter=lfs diff=lfs merge=lfs -text
+data/대기질/2023.feather filter=lfs diff=lfs merge=lfs -text
+data/대기질/air.feather filter=lfs diff=lfs merge=lfs -text
+data/대기질/air_add_위도경도.feather filter=lfs diff=lfs merge=lfs -text
+data/대기질/air_dataon_merged.feather filter=lfs diff=lfs merge=lfs -text
+data/대기질/air_dup.feather filter=lfs diff=lfs merge=lfs -text

Analysis_code/0.air_data_merge.ipynb

@@ -0,0 +1,1029 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "/opt/conda/lib/python3.8/site-packages/tqdm/auto.py:22: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
+      "  from .autonotebook import tqdm as notebook_tqdm\n"
+     ]
+    }
+   ],
+   "source": [
+    "import os\n",
+    "import numpy as np\n",
+    "import pandas as pd\n",
+    "import natsort\n",
+    "from datetime import datetime\n",
+    "from tqdm.auto import tqdm"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def get_data(year):\n",
+    "    files = natsort.natsorted(os.listdir(f'../data/대기질/{year}/'))\n",
+    "    data = []\n",
+    "    for file in tqdm(files, desc=f\"Reading files...({len(files)})\"):\n",
+    "        data.append(pd.read_excel(f'../data/대기질/{year}/{file}', usecols=[\"지역\", '망', \"측정소코드\", \"측정소명\", \"측정일시\", \"O3\", \"NO2\", \"PM10\", \"PM25\", \"주소\"]))\n",
+    "\n",
+    "    return pd.concat(data)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# 합친 데이터에 날짜 정보를 추가한다.\n",
+    "def add_date(df):\n",
+    "\n",
+    "    df[\"측정일시\"] = df[\"측정일시\"].astype(str).str[:10]\n",
+    "    df[\"측정일시\"] = pd.to_datetime(df[\"측정일시\"], format='%Y%m%d%H', errors=\"coerce\")\n",
+    "\n",
+    "    df[\"year\"] = df[\"측정일시\"].dt.year\n",
+    "    df[\"month\"] = df[\"측정일시\"].dt.month\n",
+    "    df[\"day\"] = df[\"측정일시\"].dt.day\n",
+    "    df[\"hour\"] = df[\"측정일시\"].dt.hour\n",
+    "\n",
+    "    return df"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "  0%|          | 0/6 [00:00<?, ?it/s]\n",
+      "Reading files...(13):   0%|          | 0/13 [00:00<?, ?it/s]\u001b[A\n",
+      "Reading files...(13):   8%|▊         | 1/13 [00:34<06:57, 34.80s/it]\u001b[A\n",
+      "Reading files...(13):  15%|█▌        | 2/13 [01:12<06:41, 36.47s/it]\u001b[A\n",
+      "Reading files...(13):  23%|██▎       | 3/13 [01:47<05:58, 35.89s/it]\u001b[A\n",
+      "Reading files...(13):  31%|███       | 4/13 [02:23<05:23, 35.96s/it]\u001b[A\n",
+      "Reading files...(13):  38%|███▊      | 5/13 [02:59<04:47, 35.92s/it]\u001b[A\n",
+      "Reading files...(13):  46%|████▌     | 6/13 [03:35<04:12, 36.09s/it]\u001b[A\n",
+      "Reading files...(13):  62%|██████▏   | 8/13 [04:12<02:16, 27.35s/it]\u001b[A\n",
+      "Reading files...(13):  69%|██████▉   | 9/13 [04:46<01:56, 29.05s/it]\u001b[A\n",
+      "Reading files...(13):  77%|███████▋  | 10/13 [05:21<01:31, 30.55s/it]\u001b[A\n",
+      "Reading files...(13):  85%|████████▍ | 11/13 [05:58<01:04, 32.46s/it]\u001b[A\n",
+      "Reading files...(13):  92%|█████████▏| 12/13 [06:37<00:34, 34.28s/it]\u001b[A\n",
+      "Reading files...(13): 100%|██████████| 13/13 [07:08<00:00, 32.93s/it]\u001b[A\n",
+      " 17%|█▋        | 1/6 [07:18<36:30, 438.18s/it]\n",
+      "Reading files...(13):   0%|          | 0/13 [00:00<?, ?it/s]\u001b[A\n",
+      "Reading files...(13):   8%|▊         | 1/13 [00:43<08:41, 43.43s/it]\u001b[A\n",
+      "Reading files...(13):  15%|█▌        | 2/13 [01:26<07:56, 43.29s/it]\u001b[A\n",
+      "Reading files...(13):  23%|██▎       | 3/13 [02:07<07:02, 42.22s/it]\u001b[A\n",
+      "Reading files...(13):  31%|███       | 4/13 [02:50<06:23, 42.66s/it]\u001b[A\n",
+      "Reading files...(13):  38%|███▊      | 5/13 [03:28<05:27, 40.90s/it]\u001b[A\n",
+      "Reading files...(13):  46%|████▌     | 6/13 [04:15<04:59, 42.79s/it]\u001b[A\n",
+      "Reading files...(13):  54%|█████▍    | 7/13 [04:58<04:18, 43.14s/it]\u001b[A\n",
+      "Reading files...(13):  62%|██████▏   | 8/13 [05:43<03:37, 43.47s/it]\u001b[A\n",
+      "Reading files...(13):  69%|██████▉   | 9/13 [06:28<02:55, 43.96s/it]\u001b[A\n",
+      "Reading files...(13):  77%|███████▋  | 10/13 [07:12<02:12, 44.01s/it]\u001b[A\n",
+      "Reading files...(13):  85%|████████▍ | 11/13 [07:52<01:25, 42.90s/it]\u001b[A\n",
+      "Reading files...(13): 100%|██████████| 13/13 [08:34<00:00, 39.61s/it]\u001b[A\n",
+      " 33%|███▎      | 2/6 [16:05<32:42, 490.55s/it]\n",
+      "Reading files...(13):   0%|          | 0/13 [00:00<?, ?it/s]\u001b[A\n",
+      "Reading files...(13):   8%|▊         | 1/13 [00:49<09:56, 49.74s/it]\u001b[A\n",
+      "Reading files...(13):  15%|█▌        | 2/13 [01:43<09:31, 51.98s/it]\u001b[A\n",
+      "Reading files...(13):  23%|██▎       | 3/13 [02:33<08:29, 50.96s/it]\u001b[A\n",
+      "Reading files...(13):  31%|███       | 4/13 [03:23<07:38, 50.95s/it]\u001b[A\n",
+      "Reading files...(13):  38%|███▊      | 5/13 [04:13<06:43, 50.46s/it]\u001b[A\n",
+      "Reading files...(13):  46%|████▌     | 6/13 [04:58<05:40, 48.71s/it]\u001b[A\n",
+      "Reading files...(13):  54%|█████▍    | 7/13 [05:50<04:57, 49.66s/it]\u001b[A\n",
+      "Reading files...(13):  62%|██████▏   | 8/13 [06:45<04:16, 51.29s/it]\u001b[A\n",
+      "Reading files...(13):  77%|███████▋  | 10/13 [07:38<01:58, 39.46s/it]\u001b[A\n",
+      "Reading files...(13):  85%|████████▍ | 11/13 [08:30<01:25, 42.79s/it]\u001b[A\n",
+      "Reading files...(13):  92%|█████████▏| 12/13 [09:26<00:46, 46.32s/it]\u001b[A\n",
+      "Reading files...(13): 100%|██████████| 13/13 [10:13<00:00, 47.19s/it]\u001b[A\n",
+      " 50%|█████     | 3/6 [26:32<27:38, 552.96s/it]\n",
+      "Reading files...(13):   0%|          | 0/13 [00:00<?, ?it/s]\u001b[A\n",
+      "Reading files...(13):   8%|▊         | 1/13 [00:59<11:48, 59.01s/it]\u001b[A\n",
+      "Reading files...(13):  15%|█▌        | 2/13 [01:56<10:40, 58.19s/it]\u001b[A\n",
+      "Reading files...(13):  23%|██▎       | 3/13 [02:53<09:37, 57.77s/it]\u001b[A\n",
+      "Reading files...(13):  31%|███       | 4/13 [03:52<08:41, 58.00s/it]\u001b[A\n",
+      "Reading files...(13):  38%|███▊      | 5/13 [04:44<07:26, 55.77s/it]\u001b[A\n",
+      "Reading files...(13):  46%|████▌     | 6/13 [05:40<06:32, 56.05s/it]\u001b[A\n",
+      "Reading files...(13):  54%|█████▍    | 7/13 [06:36<05:36, 56.06s/it]\u001b[A\n",
+      "Reading files...(13):  62%|██████▏   | 8/13 [07:33<04:42, 56.42s/it]\u001b[A\n",
+      "Reading files...(13):  69%|██████▉   | 9/13 [08:34<03:51, 57.76s/it]\u001b[A\n",
+      "Reading files...(13):  77%|███████▋  | 10/13 [09:35<02:56, 58.75s/it]\u001b[A\n",
+      "Reading files...(13):  92%|█████████▏| 12/13 [10:33<00:44, 44.84s/it]\u001b[A\n",
+      "Reading files...(13): 100%|██████████| 13/13 [11:32<00:00, 53.29s/it]\u001b[A\n",
+      " 67%|██████▋   | 4/6 [38:20<20:28, 614.26s/it]\n",
+      "Reading files...(13):   0%|          | 0/13 [00:00<?, ?it/s]\u001b[A\n",
+      "Reading files...(13):   8%|▊         | 1/13 [00:59<11:57, 59.79s/it]\u001b[A\n",
+      "Reading files...(13):  15%|█▌        | 2/13 [02:01<11:07, 60.67s/it]\u001b[A\n",
+      "Reading files...(13):  23%|██▎       | 3/13 [03:02<10:10, 61.02s/it]\u001b[A\n",
+      "Reading files...(13):  31%|███       | 4/13 [03:57<08:48, 58.74s/it]\u001b[A\n",
+      "Reading files...(13):  38%|███▊      | 5/13 [04:57<07:53, 59.18s/it]\u001b[A\n",
+      "Reading files...(13):  46%|████▌     | 6/13 [06:00<07:03, 60.45s/it]\u001b[A\n",
+      "Reading files...(13):  54%|█████▍    | 7/13 [07:00<06:02, 60.38s/it]\u001b[A\n",
+      "Reading files...(13):  62%|██████▏   | 8/13 [08:02<05:04, 60.85s/it]\u001b[A\n",
+      "Reading files...(13):  69%|██████▉   | 9/13 [09:04<04:04, 61.03s/it]\u001b[A\n",
+      "Reading files...(13):  77%|███████▋  | 10/13 [10:04<03:02, 60.67s/it]\u001b[A\n",
+      "Reading files...(13):  92%|█████████▏| 12/13 [11:06<00:46, 46.76s/it]\u001b[A\n",
+      "Reading files...(13): 100%|██████████| 13/13 [12:09<00:00, 56.08s/it]\u001b[A\n",
+      " 83%|████████▎ | 5/6 [50:46<11:01, 661.78s/it]\n",
+      "Reading files...(13):   0%|          | 0/13 [00:00<?, ?it/s]\u001b[A\n",
+      "Reading files...(13):   8%|▊         | 1/13 [01:03<12:46, 63.88s/it]\u001b[A\n",
+      "Reading files...(13):  15%|█▌        | 2/13 [02:08<11:50, 64.56s/it]\u001b[A\n",
+      "Reading files...(13):  23%|██▎       | 3/13 [03:10<10:32, 63.22s/it]\u001b[A\n",
+      "Reading files...(13):  31%|███       | 4/13 [04:07<09:05, 60.63s/it]\u001b[A\n",
+      "Reading files...(13):  38%|███▊      | 5/13 [05:09<08:11, 61.41s/it]\u001b[A\n",
+      "Reading files...(13):  46%|████▌     | 6/13 [06:12<07:13, 61.92s/it]\u001b[A\n",
+      "Reading files...(13):  54%|█████▍    | 7/13 [07:13<06:09, 61.50s/it]\u001b[A\n",
+      "Reading files...(13):  62%|██████▏   | 8/13 [08:15<05:08, 61.64s/it]\u001b[A\n",
+      "Reading files...(13):  69%|██████▉   | 9/13 [09:17<04:07, 61.81s/it]\u001b[A\n",
+      "Reading files...(13):  77%|███████▋  | 10/13 [10:19<03:05, 61.96s/it]\u001b[A\n",
+      "Reading files...(13):  92%|█████████▏| 12/13 [11:23<00:47, 47.75s/it]\u001b[A\n",
+      "Reading files...(13): 100%|██████████| 13/13 [12:27<00:00, 57.50s/it]\u001b[A\n",
+      "100%|██████████| 6/6 [1:03:31<00:00, 635.28s/it]\n"
+     ]
+    }
+   ],
+   "source": [
+    "import os\n",
+    "import pandas as pd\n",
+    "from tqdm.auto import tqdm\n",
+    "\n",
+    "# 대기질 데이터를 불러와서 하나의 파일로 합친다.\n",
+    "def get_data(year):\n",
+    "    directory = f'../data/대기질/{year}/'\n",
+    "    files = os.listdir(directory)\n",
+    "    data = []\n",
+    "    \n",
+    "    # 파일 목록에서 디렉토리를 제외하고 오직 Excel 파일만 처리\n",
+    "    for file in tqdm(files, desc=f\"Reading files...({len(files)})\"):\n",
+    "        file_path = os.path.join(directory, file)\n",
+    "        if os.path.isfile(file_path) and file_path.endswith(('.xls', '.xlsx')):  # Excel 파일 확장자만 허용\n",
+    "            data.append(pd.read_excel(file_path, usecols=[\"지역\", '망', \"측정소코드\", \"측정소명\", \"측정일시\", \"O3\", \"NO2\", \"PM10\", \"PM25\", \"주소\"]))\n",
+    "    \n",
+    "    return pd.concat(data)\n",
+    "\n",
+    "years = [2018, 2019, 2020,2021,2022,2023] # 2018년부터 2023년까지의 데이터를 합친다.\n",
+    "for year in tqdm(years):\n",
+    "    data = get_data(year)\n",
+    "    data = add_date(data)\n",
+    "    data.reset_index(drop=True, inplace=True)\n",
+    "    data.to_feather(f\"../data/대기질/{year}.feather\")\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>지역</th>\n",
+       "      <th>망</th>\n",
+       "      <th>측정소코드</th>\n",
+       "      <th>측정소명</th>\n",
+       "      <th>측정일시</th>\n",
+       "      <th>O3</th>\n",
+       "      <th>NO2</th>\n",
+       "      <th>PM10</th>\n",
+       "      <th>PM25</th>\n",
+       "      <th>주소</th>\n",
+       "      <th>year</th>\n",
+       "      <th>month</th>\n",
+       "      <th>day</th>\n",
+       "      <th>hour</th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "    <tr>\n",
+       "      <th>0</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 01:00:00</td>\n",
+       "      <td>0.0249</td>\n",
+       "      <td>0.0188</td>\n",
+       "      <td>21.0</td>\n",
+       "      <td>19.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>1.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 02:00:00</td>\n",
+       "      <td>0.0263</td>\n",
+       "      <td>0.0163</td>\n",
+       "      <td>18.0</td>\n",
+       "      <td>15.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>2.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>2</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 03:00:00</td>\n",
+       "      <td>0.0218</td>\n",
+       "      <td>0.0192</td>\n",
+       "      <td>24.0</td>\n",
+       "      <td>21.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>3.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>3</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 04:00:00</td>\n",
+       "      <td>0.0131</td>\n",
+       "      <td>0.0214</td>\n",
+       "      <td>25.0</td>\n",
+       "      <td>19.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>4.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>4</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 05:00:00</td>\n",
+       "      <td>0.0131</td>\n",
+       "      <td>0.0160</td>\n",
+       "      <td>25.0</td>\n",
+       "      <td>21.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>5.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>5</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 06:00:00</td>\n",
+       "      <td>0.0115</td>\n",
+       "      <td>0.0196</td>\n",
+       "      <td>23.0</td>\n",
+       "      <td>18.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>6.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>6</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 07:00:00</td>\n",
+       "      <td>0.0094</td>\n",
+       "      <td>0.0230</td>\n",
+       "      <td>26.0</td>\n",
+       "      <td>21.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>7.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>7</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 08:00:00</td>\n",
+       "      <td>0.0222</td>\n",
+       "      <td>0.0175</td>\n",
+       "      <td>26.0</td>\n",
+       "      <td>20.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>8.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>8</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 09:00:00</td>\n",
+       "      <td>0.0396</td>\n",
+       "      <td>0.0153</td>\n",
+       "      <td>27.0</td>\n",
+       "      <td>20.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>9.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>9</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 10:00:00</td>\n",
+       "      <td>0.0530</td>\n",
+       "      <td>0.0105</td>\n",
+       "      <td>19.0</td>\n",
+       "      <td>16.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>10.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>10</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 11:00:00</td>\n",
+       "      <td>0.0607</td>\n",
+       "      <td>0.0090</td>\n",
+       "      <td>20.0</td>\n",
+       "      <td>20.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>11.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>11</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 12:00:00</td>\n",
+       "      <td>0.0688</td>\n",
+       "      <td>0.0114</td>\n",
+       "      <td>20.0</td>\n",
+       "      <td>17.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>12.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>12</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 13:00:00</td>\n",
+       "      <td>0.0758</td>\n",
+       "      <td>0.0101</td>\n",
+       "      <td>23.0</td>\n",
+       "      <td>17.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>13.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>13</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 14:00:00</td>\n",
+       "      <td>0.0743</td>\n",
+       "      <td>0.0093</td>\n",
+       "      <td>20.0</td>\n",
+       "      <td>17.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>14.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>14</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 15:00:00</td>\n",
+       "      <td>0.0749</td>\n",
+       "      <td>0.0100</td>\n",
+       "      <td>19.0</td>\n",
+       "      <td>11.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>15.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>15</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 16:00:00</td>\n",
+       "      <td>0.0716</td>\n",
+       "      <td>0.0092</td>\n",
+       "      <td>19.0</td>\n",
+       "      <td>15.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>16.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>16</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 17:00:00</td>\n",
+       "      <td>0.0613</td>\n",
+       "      <td>0.0099</td>\n",
+       "      <td>18.0</td>\n",
+       "      <td>15.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>17.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>17</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 18:00:00</td>\n",
+       "      <td>0.0496</td>\n",
+       "      <td>0.0098</td>\n",
+       "      <td>18.0</td>\n",
+       "      <td>14.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>18.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>18</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 19:00:00</td>\n",
+       "      <td>0.0473</td>\n",
+       "      <td>0.0124</td>\n",
+       "      <td>17.0</td>\n",
+       "      <td>17.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>19.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>19</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 20:00:00</td>\n",
+       "      <td>0.0498</td>\n",
+       "      <td>0.0170</td>\n",
+       "      <td>17.0</td>\n",
+       "      <td>15.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>20.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>20</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 21:00:00</td>\n",
+       "      <td>0.0616</td>\n",
+       "      <td>0.0134</td>\n",
+       "      <td>23.0</td>\n",
+       "      <td>20.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>21.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>21</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 22:00:00</td>\n",
+       "      <td>0.0543</td>\n",
+       "      <td>0.0109</td>\n",
+       "      <td>18.0</td>\n",
+       "      <td>16.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>22.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>22</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-01 23:00:00</td>\n",
+       "      <td>0.0507</td>\n",
+       "      <td>0.0113</td>\n",
+       "      <td>17.0</td>\n",
+       "      <td>16.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>23.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>23</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>NaT</td>\n",
+       "      <td>0.0427</td>\n",
+       "      <td>0.0125</td>\n",
+       "      <td>17.0</td>\n",
+       "      <td>16.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>NaN</td>\n",
+       "      <td>NaN</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>24</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-02 01:00:00</td>\n",
+       "      <td>0.0334</td>\n",
+       "      <td>0.0148</td>\n",
+       "      <td>21.0</td>\n",
+       "      <td>20.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>2.0</td>\n",
+       "      <td>1.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>25</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-02 02:00:00</td>\n",
+       "      <td>0.0337</td>\n",
+       "      <td>0.0133</td>\n",
+       "      <td>22.0</td>\n",
+       "      <td>18.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>2.0</td>\n",
+       "      <td>2.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>26</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-02 03:00:00</td>\n",
+       "      <td>0.0260</td>\n",
+       "      <td>0.0162</td>\n",
+       "      <td>25.0</td>\n",
+       "      <td>20.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>2.0</td>\n",
+       "      <td>3.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>27</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-02 04:00:00</td>\n",
+       "      <td>0.0195</td>\n",
+       "      <td>0.0179</td>\n",
+       "      <td>22.0</td>\n",
+       "      <td>18.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>2.0</td>\n",
+       "      <td>4.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>28</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-02 05:00:00</td>\n",
+       "      <td>0.0171</td>\n",
+       "      <td>0.0170</td>\n",
+       "      <td>19.0</td>\n",
+       "      <td>17.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>2.0</td>\n",
+       "      <td>5.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>29</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-02 06:00:00</td>\n",
+       "      <td>0.0181</td>\n",
+       "      <td>0.0145</td>\n",
+       "      <td>14.0</td>\n",
+       "      <td>10.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>2.0</td>\n",
+       "      <td>6.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>30</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-02 07:00:00</td>\n",
+       "      <td>0.0174</td>\n",
+       "      <td>0.0156</td>\n",
+       "      <td>11.0</td>\n",
+       "      <td>10.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>2.0</td>\n",
+       "      <td>7.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>31</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-02 08:00:00</td>\n",
+       "      <td>0.0213</td>\n",
+       "      <td>0.0147</td>\n",
+       "      <td>12.0</td>\n",
+       "      <td>9.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>2.0</td>\n",
+       "      <td>8.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>32</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-02 09:00:00</td>\n",
+       "      <td>0.0267</td>\n",
+       "      <td>0.0143</td>\n",
+       "      <td>11.0</td>\n",
+       "      <td>10.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>2.0</td>\n",
+       "      <td>9.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>33</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-02 10:00:00</td>\n",
+       "      <td>0.0289</td>\n",
+       "      <td>0.0155</td>\n",
+       "      <td>12.0</td>\n",
+       "      <td>9.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>2.0</td>\n",
+       "      <td>10.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>34</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-02 11:00:00</td>\n",
+       "      <td>0.0381</td>\n",
+       "      <td>0.0108</td>\n",
+       "      <td>13.0</td>\n",
+       "      <td>13.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>2.0</td>\n",
+       "      <td>11.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>35</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-02 12:00:00</td>\n",
+       "      <td>0.0441</td>\n",
+       "      <td>0.0079</td>\n",
+       "      <td>13.0</td>\n",
+       "      <td>12.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>2.0</td>\n",
+       "      <td>12.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>36</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-02 13:00:00</td>\n",
+       "      <td>0.0489</td>\n",
+       "      <td>0.0067</td>\n",
+       "      <td>8.0</td>\n",
+       "      <td>10.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>2.0</td>\n",
+       "      <td>13.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>37</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-02 14:00:00</td>\n",
+       "      <td>0.0498</td>\n",
+       "      <td>0.0072</td>\n",
+       "      <td>11.0</td>\n",
+       "      <td>10.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>2.0</td>\n",
+       "      <td>14.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>38</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-02 15:00:00</td>\n",
+       "      <td>0.0459</td>\n",
+       "      <td>0.0073</td>\n",
+       "      <td>14.0</td>\n",
+       "      <td>12.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>2.0</td>\n",
+       "      <td>15.0</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>39</th>\n",
+       "      <td>서울 중구</td>\n",
+       "      <td>도시대기</td>\n",
+       "      <td>111121</td>\n",
+       "      <td>중구</td>\n",
+       "      <td>2023-07-02 16:00:00</td>\n",
+       "      <td>0.0474</td>\n",
+       "      <td>0.0079</td>\n",
+       "      <td>12.0</td>\n",
+       "      <td>11.0</td>\n",
+       "      <td>서울 중구 덕수궁길 15</td>\n",
+       "      <td>2023.0</td>\n",
+       "      <td>7.0</td>\n",
+       "      <td>2.0</td>\n",
+       "      <td>16.0</td>\n",
+       "    </tr>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "       지역     망   측정소코드 측정소명                측정일시      O3     NO2  PM10  PM25  \\\n",
+       "0   서울 중구  도시대기  111121   중구 2023-07-01 01:00:00  0.0249  0.0188  21.0  19.0   \n",
+       "1   서울 중구  도시대기  111121   중구 2023-07-01 02:00:00  0.0263  0.0163  18.0  15.0   \n",
+       "2   서울 중구  도시대기  111121   중구 2023-07-01 03:00:00  0.0218  0.0192  24.0  21.0   \n",
+       "3   서울 중구  도시대기  111121   중구 2023-07-01 04:00:00  0.0131  0.0214  25.0  19.0   \n",
+       "4   서울 중구  도시대기  111121   중구 2023-07-01 05:00:00  0.0131  0.0160  25.0  21.0   \n",
+       "5   서울 중구  도시대기  111121   중구 2023-07-01 06:00:00  0.0115  0.0196  23.0  18.0   \n",
+       "6   서울 중구  도시대기  111121   중구 2023-07-01 07:00:00  0.0094  0.0230  26.0  21.0   \n",
+       "7   서울 중구  도시대기  111121   중구 2023-07-01 08:00:00  0.0222  0.0175  26.0  20.0   \n",
+       "8   서울 중구  도시대기  111121   중구 2023-07-01 09:00:00  0.0396  0.0153  27.0  20.0   \n",
+       "9   서울 중구  도시대기  111121   중구 2023-07-01 10:00:00  0.0530  0.0105  19.0  16.0   \n",
+       "10  서울 중구  도시대기  111121   중구 2023-07-01 11:00:00  0.0607  0.0090  20.0  20.0   \n",
+       "11  서울 중구  도시대기  111121   중구 2023-07-01 12:00:00  0.0688  0.0114  20.0  17.0   \n",
+       "12  서울 중구  도시대기  111121   중구 2023-07-01 13:00:00  0.0758  0.0101  23.0  17.0   \n",
+       "13  서울 중구  도시대기  111121   중구 2023-07-01 14:00:00  0.0743  0.0093  20.0  17.0   \n",
+       "14  서울 중구  도시대기  111121   중구 2023-07-01 15:00:00  0.0749  0.0100  19.0  11.0   \n",
+       "15  서울 중구  도시대기  111121   중구 2023-07-01 16:00:00  0.0716  0.0092  19.0  15.0   \n",
+       "16  서울 중구  도시대기  111121   중구 2023-07-01 17:00:00  0.0613  0.0099  18.0  15.0   \n",
+       "17  서울 중구  도시대기  111121   중구 2023-07-01 18:00:00  0.0496  0.0098  18.0  14.0   \n",
+       "18  서울 중구  도시대기  111121   중구 2023-07-01 19:00:00  0.0473  0.0124  17.0  17.0   \n",
+       "19  서울 중구  도시대기  111121   중구 2023-07-01 20:00:00  0.0498  0.0170  17.0  15.0   \n",
+       "20  서울 중구  도시대기  111121   중구 2023-07-01 21:00:00  0.0616  0.0134  23.0  20.0   \n",
+       "21  서울 중구  도시대기  111121   중구 2023-07-01 22:00:00  0.0543  0.0109  18.0  16.0   \n",
+       "22  서울 중구  도시대기  111121   중구 2023-07-01 23:00:00  0.0507  0.0113  17.0  16.0   \n",
+       "23  서울 중구  도시대기  111121   중구                 NaT  0.0427  0.0125  17.0  16.0   \n",
+       "24  서울 중구  도시대기  111121   중구 2023-07-02 01:00:00  0.0334  0.0148  21.0  20.0   \n",
+       "25  서울 중구  도시대기  111121   중구 2023-07-02 02:00:00  0.0337  0.0133  22.0  18.0   \n",
+       "26  서울 중구  도시대기  111121   중구 2023-07-02 03:00:00  0.0260  0.0162  25.0  20.0   \n",
+       "27  서울 중구  도시대기  111121   중구 2023-07-02 04:00:00  0.0195  0.0179  22.0  18.0   \n",
+       "28  서울 중구  도시대기  111121   중구 2023-07-02 05:00:00  0.0171  0.0170  19.0  17.0   \n",
+       "29  서울 중구  도시대기  111121   중구 2023-07-02 06:00:00  0.0181  0.0145  14.0  10.0   \n",
+       "30  서울 중구  도시대기  111121   중구 2023-07-02 07:00:00  0.0174  0.0156  11.0  10.0   \n",
+       "31  서울 중구  도시대기  111121   중구 2023-07-02 08:00:00  0.0213  0.0147  12.0   9.0   \n",
+       "32  서울 중구  도시대기  111121   중구 2023-07-02 09:00:00  0.0267  0.0143  11.0  10.0   \n",
+       "33  서울 중구  도시대기  111121   중구 2023-07-02 10:00:00  0.0289  0.0155  12.0   9.0   \n",
+       "34  서울 중구  도시대기  111121   중구 2023-07-02 11:00:00  0.0381  0.0108  13.0  13.0   \n",
+       "35  서울 중구  도시대기  111121   중구 2023-07-02 12:00:00  0.0441  0.0079  13.0  12.0   \n",
+       "36  서울 중구  도시대기  111121   중구 2023-07-02 13:00:00  0.0489  0.0067   8.0  10.0   \n",
+       "37  서울 중구  도시대기  111121   중구 2023-07-02 14:00:00  0.0498  0.0072  11.0  10.0   \n",
+       "38  서울 중구  도시대기  111121   중구 2023-07-02 15:00:00  0.0459  0.0073  14.0  12.0   \n",
+       "39  서울 중구  도시대기  111121   중구 2023-07-02 16:00:00  0.0474  0.0079  12.0  11.0   \n",
+       "\n",
+       "               주소    year  month  day  hour  \n",
+       "0   서울 중구 덕수궁길 15  2023.0    7.0  1.0   1.0  \n",
+       "1   서울 중구 덕수궁길 15  2023.0    7.0  1.0   2.0  \n",
+       "2   서울 중구 덕수궁길 15  2023.0    7.0  1.0   3.0  \n",
+       "3   서울 중구 덕수궁길 15  2023.0    7.0  1.0   4.0  \n",
+       "4   서울 중구 덕수궁길 15  2023.0    7.0  1.0   5.0  \n",
+       "5   서울 중구 덕수궁길 15  2023.0    7.0  1.0   6.0  \n",
+       "6   서울 중구 덕수궁길 15  2023.0    7.0  1.0   7.0  \n",
+       "7   서울 중구 덕수궁길 15  2023.0    7.0  1.0   8.0  \n",
+       "8   서울 중구 덕수궁길 15  2023.0    7.0  1.0   9.0  \n",
+       "9   서울 중구 덕수궁길 15  2023.0    7.0  1.0  10.0  \n",
+       "10  서울 중구 덕수궁길 15  2023.0    7.0  1.0  11.0  \n",
+       "11  서울 중구 덕수궁길 15  2023.0    7.0  1.0  12.0  \n",
+       "12  서울 중구 덕수궁길 15  2023.0    7.0  1.0  13.0  \n",
+       "13  서울 중구 덕수궁길 15  2023.0    7.0  1.0  14.0  \n",
+       "14  서울 중구 덕수궁길 15  2023.0    7.0  1.0  15.0  \n",
+       "15  서울 중구 덕수궁길 15  2023.0    7.0  1.0  16.0  \n",
+       "16  서울 중구 덕수궁길 15  2023.0    7.0  1.0  17.0  \n",
+       "17  서울 중구 덕수궁길 15  2023.0    7.0  1.0  18.0  \n",
+       "18  서울 중구 덕수궁길 15  2023.0    7.0  1.0  19.0  \n",
+       "19  서울 중구 덕수궁길 15  2023.0    7.0  1.0  20.0  \n",
+       "20  서울 중구 덕수궁길 15  2023.0    7.0  1.0  21.0  \n",
+       "21  서울 중구 덕수궁길 15  2023.0    7.0  1.0  22.0  \n",
+       "22  서울 중구 덕수궁길 15  2023.0    7.0  1.0  23.0  \n",
+       "23  서울 중구 덕수궁길 15     NaN    NaN  NaN   NaN  \n",
+       "24  서울 중구 덕수궁길 15  2023.0    7.0  2.0   1.0  \n",
+       "25  서울 중구 덕수궁길 15  2023.0    7.0  2.0   2.0  \n",
+       "26  서울 중구 덕수궁길 15  2023.0    7.0  2.0   3.0  \n",
+       "27  서울 중구 덕수궁길 15  2023.0    7.0  2.0   4.0  \n",
+       "28  서울 중구 덕수궁길 15  2023.0    7.0  2.0   5.0  \n",
+       "29  서울 중구 덕수궁길 15  2023.0    7.0  2.0   6.0  \n",
+       "30  서울 중구 덕수궁길 15  2023.0    7.0  2.0   7.0  \n",
+       "31  서울 중구 덕수궁길 15  2023.0    7.0  2.0   8.0  \n",
+       "32  서울 중구 덕수궁길 15  2023.0    7.0  2.0   9.0  \n",
+       "33  서울 중구 덕수궁길 15  2023.0    7.0  2.0  10.0  \n",
+       "34  서울 중구 덕수궁길 15  2023.0    7.0  2.0  11.0  \n",
+       "35  서울 중구 덕수궁길 15  2023.0    7.0  2.0  12.0  \n",
+       "36  서울 중구 덕수궁길 15  2023.0    7.0  2.0  13.0  \n",
+       "37  서울 중구 덕수궁길 15  2023.0    7.0  2.0  14.0  \n",
+       "38  서울 중구 덕수궁길 15  2023.0    7.0  2.0  15.0  \n",
+       "39  서울 중구 덕수궁길 15  2023.0    7.0  2.0  16.0  "
+      ]
+     },
+     "execution_count": 6,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "data.head(40)"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.8.13"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}

Analysis_code/2.eda_preproccesing.ipynb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8fe49fd26e48bcada89076a0b3c8ffe45e3d2e8407fe953b1558df9bfcfcddb
+size 41518580

Analysis_code/best_deepgbm_model.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:471ce03d818bdefc9631121537afd4b771d85d872f72d74634e5cee824de2b62
+size 988522

Analysis_code/best_model_f1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12c6e86e039528b07c4cc6e90e3033da7b02d67c33539cbb81d280da281f46c3
+size 8999933

Analysis_code/deepgbm.py

@@ -0,0 +1,47 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+class DeepGBM(nn.Module):
+    def __init__(self, num_features, cat_features, num_classes, d_main=128, d_hidden=64, n_blocks=4, dropout=0.2):
+        super(DeepGBM, self).__init__()
+
+        self.num_classes = num_classes
+        
+        # 연속형 변수 처리 (Linear)
+        self.num_linear = nn.Linear(num_features, d_main)
+
+        # 범주형 변수 처리 (Embedding)
+        self.cat_embedding = nn.ModuleList([
+            nn.Embedding(cat_size, d_main) for cat_size in cat_features
+        ])
+
+        # ResNet-like 블록
+        self.blocks = nn.ModuleList([
+            nn.Sequential(
+                nn.Linear(d_main, d_hidden),
+                nn.ReLU(),
+                nn.Dropout(dropout),
+                nn.Linear(d_hidden, d_main),
+                nn.ReLU()
+            ) for _ in range(n_blocks)
+        ])
+
+        if num_classes == 2:
+            self.output_layer = nn.Linear(d_main, 1)  # Binary classification
+        elif num_classes > 2:
+            self.output_layer = nn.Linear(d_main, num_classes)  # Multi classification
+
+    def forward(self, x_num, x_cat):  # 두 개의 입력을 받음
+        x_num = self.num_linear(x_num)
+
+        # 범주형 변수를 임베딩 후 합산
+        x_cat = [embed(x_cat[:, i]) for i, embed in enumerate(self.cat_embedding)]
+        x_cat = torch.stack(x_cat, dim=1).sum(dim=1)  
+        x = x_num + x_cat  # 연속형 + 범주형 결합
+        
+        for block in self.blocks:
+            x = x + block(x)  # Residual connection
+        x = self.output_layer(x)
+        
+        return x

Analysis_code/final_test/final.ipynb

@@ -0,0 +1,1143 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import pandas as pd\n",
+    "import numpy as np\n",
+    "from sklearn.preprocessing import StandardScaler, LabelEncoder\n",
+    "import torch\n",
+    "from torch.utils.data import DataLoader, TensorDataset\n",
+    "import random\n",
+    "from collections import Counter\n",
+    "import sys\n",
+    "sys.path.append('../../../../../../../../mnt/workspace/LightGBM/python-package')\n",
+    "from lightgbm import LGBMClassifier\n",
+    "import numpy as np\n",
+    "from sklearn.model_selection import train_test_split\n",
+    "from sklearn.inspection import permutation_importance\n",
+    "from sklearn.metrics import confusion_matrix,accuracy_score, recall_score, precision_score\n",
+    "from sklearn.model_selection import StratifiedKFold\n",
+    "from xgboost import XGBClassifier\n",
+    "from warnings import filterwarnings\n",
+    "filterwarnings('ignore')\n",
+    "import sys\n",
+    "sys.path.append('../')\n",
+    "import torch\n",
+    "import torch.nn as nn\n",
+    "import torch.optim as optim\n",
+    "import optuna\n",
+    "import pandas as pd\n",
+    "import numpy as np\n",
+    "import random\n",
+    "from ft_transformer import FTTransformer\n",
+    "from resnet_like import ResNetLike\n",
+    "from deepgbm import DeepGBM\n",
+    "from pytorch_tabnet.tab_model import TabNetClassifier\n",
+    "from sklearn.metrics import accuracy_score, f1_score, roc_auc_score, confusion_matrix"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Python 및 Numpy 시드 고정\n",
+    "seed = 42\n",
+    "random.seed(seed)\n",
+    "np.random.seed(seed)\n",
+    "\n",
+    "# PyTorch 시드 고정\n",
+    "torch.manual_seed(seed)\n",
+    "torch.cuda.manual_seed(seed)\n",
+    "torch.cuda.manual_seed_all(seed)  # Multi-GPU 환경에서 동일한 시드 적용\n",
+    "\n",
+    "# PyTorch 연산의 결정적 모드 설정\n",
+    "torch.backends.cudnn.deterministic = True  # 실행마다 동일한 결과를 보장\n",
+    "torch.backends.cudnn.benchmark = True  # 성능 최적화를 활성화 (가능한 한 빠른 연산 수행)\n",
+    "\n",
+    "# 전처리 함수\n",
+    "def preprocessing(df):\n",
+    "    df = df[df.columns].copy()\n",
+    "    df.loc[df['wind_dir']=='정온', 'wind_dir'] = \"0\"\n",
+    "    df['wind_dir'] = df['wind_dir'].astype('int')\n",
+    "    df['lm_cloudcover'] = df['lm_cloudcover'].astype('int')\n",
+    "    df['cloudcover'] = df['cloudcover'].astype('int')\n",
+    "    return df\n",
+    "\n",
+    "# 데이터셋 준비 함수\n",
+    "def prepare_dataset(region, data_sample='pure', target='multi', fold=3):\n",
+    "\n",
+    "    # 데이터 경로 지정\n",
+    "    dat_path = f\"../../data/data_for_modeling/{region}_train.csv\"\n",
+    "    if data_sample == 'pure':\n",
+    "        train_path = dat_path\n",
+    "    else:\n",
+    "        train_path = f'../../data/data_oversampled/{data_sample}/{data_sample}_{fold}_{region}.csv'\n",
+    "        train_path = f'../../data/data_oversampled/{data_sample}/{data_sample}_{fold}_{region}.csv'\n",
+    "    test_path = f\"../../data/data_for_modeling/{region}_test.csv\"\n",
+    "    drop_col = ['binary_class','multi_class','visi','year']\n",
+    "    target_col = f'{target}_class'\n",
+    "    \n",
+    "    # 데이터 로드\n",
+    "    region_dat = preprocessing(pd.read_csv(dat_path, index_col=0))\n",
+    "    if data_sample == 'pure':\n",
+    "        region_train = region_dat.loc[~region_dat['year'].isin([2021-fold]), :]\n",
+    "    else:\n",
+    "        region_train = preprocessing(pd.read_csv(train_path))\n",
+    "    region_val = region_dat.loc[region_dat['year'].isin([2021-fold]), :]\n",
+    "    region_test = preprocessing(pd.read_csv(test_path))\n",
+    "\n",
+    "    # 컬럼 정렬 (일관성 유지)\n",
+    "    common_columns = region_train.columns.to_list()\n",
+    "    train_data = region_train[common_columns]\n",
+    "    val_data = region_val[common_columns]\n",
+    "    test_data = region_test[common_columns]\n",
+    "\n",
+    "    # 설명변수 & 타겟 분리\n",
+    "    X_train = train_data.drop(columns=drop_col)\n",
+    "    y_train = train_data[target_col]\n",
+    "    X_val = val_data.drop(columns=drop_col)\n",
+    "    y_val = val_data[target_col]\n",
+    "    X_test = test_data.drop(columns=drop_col)\n",
+    "    y_test = test_data[target_col]\n",
+    "\n",
+    "    # 범주형 & 연속형 변수 분리\n",
+    "    categorical_cols = X_train.select_dtypes(include=['object', 'category', 'int64']).columns\n",
+    "    numerical_cols = X_train.select_dtypes(include=['float64']).columns\n",
+    "\n",
+    "    # 범주형 변수 Label Encoding\n",
+    "    label_encoders = {}\n",
+    "    for col in categorical_cols:\n",
+    "        le = LabelEncoder()\n",
+    "        le.fit(X_train[col])  # Train 데이터 기준으로 학���\n",
+    "        label_encoders[col] = le\n",
+    "\n",
+    "    # 변환 적용\n",
+    "    for col in categorical_cols:\n",
+    "        X_train[col] = label_encoders[col].transform(X_train[col])\n",
+    "        X_val[col] = label_encoders[col].transform(X_val[col])\n",
+    "        X_test[col] = label_encoders[col].transform(X_test[col])\n",
+    "\n",
+    "    # 연속형 변수 Standard Scaling\n",
+    "    scaler = StandardScaler()\n",
+    "    scaler.fit(X_train[numerical_cols])  # Train 데이터 기준으로 학습\n",
+    "\n",
+    "    # 변환 적용\n",
+    "    X_train[numerical_cols] = scaler.transform(X_train[numerical_cols])\n",
+    "    X_val[numerical_cols] = scaler.transform(X_val[numerical_cols])\n",
+    "    X_test[numerical_cols] = scaler.transform(X_test[numerical_cols])\n",
+    "\n",
+    "    return X_train, X_val, X_test, y_train, y_val, y_test, categorical_cols, numerical_cols\n",
+    "\n",
+    "# 데이터 변환 및 dataloader 생성 함수\n",
+    "def prepare_dataloader(region, data_sample='pure', target='multi', fold=3, random_state=None):\n",
+    "\n",
+    "    # 데이터 경로 지정\n",
+    "    dat_path = f\"../../data/data_for_modeling/{region}_train.csv\"\n",
+    "    if data_sample == 'pure':\n",
+    "        train_path = dat_path\n",
+    "    else:\n",
+    "        train_path = f'../../data/data_oversampled/{data_sample}/{data_sample}_{fold}_{region}.csv'\n",
+    "        train_path = f'../../data/data_oversampled/{data_sample}/{data_sample}_{fold}_{region}.csv'\n",
+    "    test_path = f\"../../data/data_for_modeling/{region}_test.csv\"\n",
+    "    drop_col = ['binary_class','multi_class','visi','year']\n",
+    "    target_col = f'{target}_class'\n",
+    "    \n",
+    "    # 데이터 로드\n",
+    "    region_dat = preprocessing(pd.read_csv(dat_path, index_col=0))\n",
+    "    if data_sample == 'pure':\n",
+    "        region_train = region_dat.loc[~region_dat['year'].isin([2021-fold]), :]\n",
+    "    else:\n",
+    "        region_train = preprocessing(pd.read_csv(train_path))\n",
+    "    region_val = region_dat.loc[region_dat['year'].isin([2021-fold]), :]\n",
+    "    region_test = preprocessing(pd.read_csv(test_path))\n",
+    "\n",
+    "    # 컬럼 정렬 (일관성 유지)\n",
+    "    common_columns = region_train.columns.to_list()\n",
+    "    train_data = region_train[common_columns]\n",
+    "    val_data = region_val[common_columns]\n",
+    "    test_data = region_test[common_columns]\n",
+    "\n",
+    "    # 설명변수 & 타겟 분리\n",
+    "    X_train = train_data.drop(columns=drop_col)\n",
+    "    y_train = train_data[target_col]\n",
+    "    X_val = val_data.drop(columns=drop_col)\n",
+    "    y_val = val_data[target_col]\n",
+    "    X_test = test_data.drop(columns=drop_col)\n",
+    "    y_test = test_data[target_col]\n",
+    "\n",
+    "    # 범주형 & 연속형 변수 분리\n",
+    "    categorical_cols = X_train.select_dtypes(include=['object', 'category', 'int64']).columns\n",
+    "    numerical_cols = X_train.select_dtypes(include=['float64']).columns\n",
+    "\n",
+    "    # 범주형 변수 Label Encoding\n",
+    "    label_encoders = {}\n",
+    "    for col in categorical_cols:\n",
+    "        le = LabelEncoder()\n",
+    "        le.fit(X_train[col])  # Train 데이터 기준으로 학습\n",
+    "        label_encoders[col] = le\n",
+    "\n",
+    "    # 변환 적용\n",
+    "    for col in categorical_cols:\n",
+    "        X_train[col] = label_encoders[col].transform(X_train[col])\n",
+    "        X_val[col] = label_encoders[col].transform(X_val[col])\n",
+    "        X_test[col] = label_encoders[col].transform(X_test[col])\n",
+    "\n",
+    "    # 연속형 변수 Standard Scaling\n",
+    "    scaler = StandardScaler()\n",
+    "    scaler.fit(X_train[numerical_cols])  # Train 데이터 기준으로 학습\n",
+    "\n",
+    "    # 변환 적용\n",
+    "    X_train[numerical_cols] = scaler.transform(X_train[numerical_cols])\n",
+    "    X_val[numerical_cols] = scaler.transform(X_val[numerical_cols])\n",
+    "    X_test[numerical_cols] = scaler.transform(X_test[numerical_cols])\n",
+    "\n",
+    "    # 연속형 변수와 범주형 변수 분리\n",
+    "    X_train_num = torch.tensor(X_train[numerical_cols].values, dtype=torch.float32)\n",
+    "    X_train_cat = torch.tensor(X_train[categorical_cols].values, dtype=torch.long)\n",
+    "\n",
+    "    X_val_num = torch.tensor(X_val[numerical_cols].values, dtype=torch.float32)\n",
+    "    X_val_cat = torch.tensor(X_val[categorical_cols].values, dtype=torch.long)\n",
+    "\n",
+    "    X_test_num = torch.tensor(X_test[numerical_cols].values, dtype=torch.float32)\n",
+    "    X_test_cat = torch.tensor(X_test[categorical_cols].values, dtype=torch.long)\n",
+    "\n",
+    "    # 레이블 변환\n",
+    "    if target == \"binary\":\n",
+    "        y_train_tensor = torch.tensor(y_train.values, dtype=torch.float32)  # 이진 분류 → float32\n",
+    "        y_val_tensor = torch.tensor(y_val.values, dtype=torch.float32)\n",
+    "        y_test_tensor = torch.tensor(y_test.values, dtype=torch.float32)\n",
+    "    elif target == \"multi\":\n",
+    "        y_train_tensor = torch.tensor(y_train.values, dtype=torch.long)  # 다중 분류 → long\n",
+    "        y_val_tensor = torch.tensor(y_val.values, dtype=torch.long)\n",
+    "        y_test_tensor = torch.tensor(y_test.values, dtype=torch.long)\n",
+    "    else:\n",
+    "        raise ValueError(\"target must be 'binary' or 'multi'\")\n",
+    "\n",
+    "    # TensorDataset 생성\n",
+    "    train_dataset = TensorDataset(X_train_num, X_train_cat, y_train_tensor)\n",
+    "    val_dataset = TensorDataset(X_val_num, X_val_cat, y_val_tensor)\n",
+    "    test_dataset = TensorDataset(X_test_num, X_test_cat, y_test_tensor)\n",
+    "\n",
+    "    # DataLoader 생성\n",
+    "    if random_state == None:\n",
+    "        train_loader = DataLoader(train_dataset, batch_size=64, shuffle=True)\n",
+    "    else:\n",
+    "        train_loader = DataLoader(train_dataset, batch_size=64, shuffle=True, generator=torch.Generator().manual_seed(random_state))\n",
+    "    val_loader = DataLoader(val_dataset, batch_size=64, shuffle=False)\n",
+    "    test_loader = DataLoader(test_dataset, batch_size=64, shuffle=False)\n",
+    "    \n",
+    "    return X_train, categorical_cols, numerical_cols, train_loader, val_loader, test_loader"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "import torch\n",
+    "# 디바이스 설정 (CUDA 사용 가능하면 GPU로, 아니면 CPU로)\n",
+    "import glob\n",
+    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
+    "\n",
+    "def calculate_csi(Y_test, pred):\n",
+    "\n",
+    "    cm = confusion_matrix(Y_test, pred)  # 변수 이름을 cm으로 변경\n",
+    "    # 혼동 행렬에서 H, F, M 추출\n",
+    "    H = (cm[0, 0] + cm[1, 1])\n",
+    "    \n",
+    "    F = (cm[1, 0] + cm[2, 0] +\n",
+    "         cm[0, 1] + cm[2, 1])\n",
+    "\n",
+    "    M = (cm[0, 2] + cm[1, 2])\n",
+    "    \n",
+    "    # CSI 계산\n",
+    "    CSI = H / (H + F + M + 1e-10)\n",
+    "    return CSI\n",
+    "\n",
+    "def csi_metric(y_true, pred):\n",
+    "    y_pred_binary = np.argmax(pred, axis=1)\n",
+    "    score = calculate_csi(y_true, y_pred_binary)\n",
+    "    return 'CSI', score, True  # higher_better=True\n",
+    "\n",
+    "\n",
+    "def eval_metric_csi(y_true, pred_prob):\n",
+    "\n",
+    "    pred = np.argmax(pred_prob, axis=1)\n",
+    "    y_true = y_true\n",
+    "    y_pred = pred\n",
+    "    csi = calculate_csi(y_true, y_pred)\n",
+    "    return -1*csi\n",
+    "\n",
+    "\n",
+    "from sklearn.metrics import matthews_corrcoef, accuracy_score\n",
+    "\n",
+    "def multiclass_mcc(y_val, y_pred):\n",
+    "    \"\"\"\n",
+    "    다중 분류에서도 sklearn의 matthews_corrcoef를 그대로 사용할 수 있음.\n",
+    "    \"\"\"\n",
+    "    return matthews_corrcoef(y_val, y_pred)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import os\n",
+    "import torch\n",
+    "# 디바이스 설정 (CUDA 사용 가능하면 GPU로, 아니면 CPU로)\n",
+    "import glob\n",
+    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
+    "import warnings\n",
+    "warnings.filterwarnings('ignore')\n",
+    "\n",
+    "def calculate_csi(Y_test, pred):\n",
+    "\n",
+    "    cm = confusion_matrix(Y_test, pred)  # 변수 이름을 cm으로 변경\n",
+    "    # 혼동 행렬에서 H, F, M 추출\n",
+    "    H = (cm[0, 0] + cm[1, 1])\n",
+    "    \n",
+    "    F = (cm[1, 0] + cm[2, 0] +\n",
+    "         cm[0, 1] + cm[2, 1])\n",
+    "\n",
+    "    M = (cm[0, 2] + cm[1, 2])\n",
+    "    \n",
+    "    # CSI 계산\n",
+    "    CSI = H / (H + F + M + 1e-10)\n",
+    "    return CSI\n",
+    "\n",
+    "# Soft Voting 앙상블\n",
+    "def get_proba(region, model_choose, data_sample, fold, target='multi'):\n",
+    "    _, _, _, _,val_loader , test_loader = prepare_dataloader(region=region, data_sample=data_sample, target=target,fold=fold ,random_state=120)\n",
+    "\n",
+    "    folder_path = f'../save_model/{model_choose}/{data_sample}'\n",
+    "    model_paths = [path for path in glob.glob(f'{folder_path}/*.pth') if f'{region}' in path]\n",
+    "\n",
+    "    model = torch.load(model_paths[fold-1], weights_only=False).to(device)\n",
+    "    model.eval()\n",
+    "\n",
+    "    test_preds = []\n",
+    "\n",
+    "\n",
+    "    with torch.no_grad():\n",
+    "        for x_num_batch, x_cat_batch, _ in test_loader:\n",
+    "            output = model(x_num_batch.to(device), x_cat_batch.to(device))\n",
+    "            output = torch.softmax(output, dim=1)\n",
+    "            test_preds.extend(output.cpu().numpy())\n",
+    "\n",
+    "\n",
+    "    return test_preds\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df_seoul = pd.read_csv(\"../../data/data_for_modeling/seoul_train.csv\")\n",
+    "df_seoul_test = pd.read_csv(\"../../data/data_for_modeling/seoul_test.csv\")\n",
+    "\n",
+    "df_busan = pd.read_csv(\"../../data/data_for_modeling/busan_train.csv\")\n",
+    "df_busan_test = pd.read_csv(\"../../data/data_for_modeling/busan_test.csv\")\n",
+    "\n",
+    "df_daegu = pd.read_csv(\"../../data/data_for_modeling/daegu_train.csv\")\n",
+    "df_daegu_test = pd.read_csv(\"../../data/data_for_modeling/daegu_test.csv\")\n",
+    "\n",
+    "df_daejeon = pd.read_csv(\"../../data/data_for_modeling/daejeon_train.csv\")\n",
+    "df_daejeon_test = pd.read_csv(\"../../data/data_for_modeling/daejeon_test.csv\")\n",
+    "\n",
+    "df_incheon = pd.read_csv(\"../../data/data_for_modeling/incheon_train.csv\")\n",
+    "df_incheon_test = pd.read_csv(\"../../data/data_for_modeling/incheon_test.csv\")\n",
+    "\n",
+    "df_gwangju = pd.read_csv(\"../../data/data_for_modeling/gwangju_train.csv\")\n",
+    "df_gwangju_test = pd.read_csv(\"../../data/data_for_modeling/gwangju_test.csv\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "def preprocessing_df(df):\n",
+    "    df = df[df.columns].copy()\n",
+    "    df['year'] = df['year'].astype('int')\n",
+    "    df['month'] = df['month'].astype('int')\n",
+    "    df['hour'] = df['hour'].astype('int')\n",
+    "    df['binary_class'] = df['binary_class'].astype('int')\n",
+    "    df['multi_class'] = df['multi_class'].astype('int')\n",
+    "\n",
+    "    df.loc[df['wind_dir']=='정온', 'wind_dir'] = \"0\"\n",
+    "    df['wind_dir'] = df['wind_dir'].astype('int')\n",
+    "    df= df[['temp_C', 'precip_mm', 'wind_speed', 'wind_dir', 'hm',\n",
+    "       'vap_pressure', 'dewpoint_C', 'loc_pressure', 'sea_pressure',\n",
+    "       'solarRad', 'snow_cm', 'cloudcover', 'lm_cloudcover', 'low_cloudbase',\n",
+    "       'groundtemp', 'O3', 'NO2', 'PM10', 'PM25', 'year',\n",
+    "       'month', 'hour', 'ground_temp - temp_C', 'hour_sin', 'hour_cos',\n",
+    "       'month_sin', 'month_cos','multi_class']]\n",
+    "    return df\n",
+    "\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df_seoul_test= preprocessing_df(df_seoul_test).copy()\n",
+    "df_busan_test= preprocessing_df(df_busan_test).copy()\n",
+    "df_daegu_test= preprocessing_df(df_daegu_test).copy()\n",
+    "df_gwangju_test= preprocessing_df(df_gwangju_test).copy()\n",
+    "df_daejeon_test= preprocessing_df(df_daejeon_test).copy()\n",
+    "df_incheon_test= preprocessing_df(df_incheon_test).copy()\n",
+    "\n",
+    "df_seoul= preprocessing_df(df_seoul).copy()\n",
+    "df_busan= preprocessing_df(df_busan).copy()\n",
+    "df_daegu= preprocessing_df(df_daegu).copy()\n",
+    "df_gwangju= preprocessing_df(df_gwangju).copy()\n",
+    "df_daejeon= preprocessing_df(df_daejeon).copy()\n",
+    "df_incheon= preprocessing_df(df_incheon).copy()\n",
+    "\n",
+    "df_seoul_test.drop(columns=['year'], inplace=True)\n",
+    "df_busan_test.drop(columns=['year'], inplace=True)\n",
+    "df_daegu_test.drop(columns=['year'], inplace=True)\n",
+    "df_daejeon_test.drop(columns=['year'], inplace=True)\n",
+    "df_incheon_test.drop(columns=['year'], inplace=True)\n",
+    "df_gwangju_test.drop(columns=['year'], inplace=True)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import joblib\n",
+    "\n",
+    "lgb_seoul= joblib.load('../save_model/LGB_optima/lgb_seoul_smote.pkl')\n",
+    "lgb_busan= joblib.load('../save_model/LGB_optima/lgb_busan_smote.pkl')\n",
+    "lgb_incheon= joblib.load('../save_model/LGB_optima/lgb_incheon_smote.pkl')\n",
+    "lgb_daegu= joblib.load('../save_model/LGB_optima/lgb_daegu_smote.pkl')\n",
+    "lgb_daejeon= joblib.load('../save_model/LGB_optima/lgb_daejeon_smote.pkl')\n",
+    "lgb_gwangju= joblib.load('../save_model/LGB_optima/lgb_gwangju_smote.pkl')\n",
+    "\n",
+    "xgb_seoul= joblib.load('../save_model/XGB_optima/xgb_seoul_smote.pkl')\n",
+    "xgb_busan= joblib.load('../save_model/XGB_optima/xgb_busan_ctgan20000.pkl')\n",
+    "xgb_incheon= joblib.load('../save_model/XGB_optima/xgb_incheon_smote.pkl')\n",
+    "xgb_daegu= joblib.load('../save_model/XGB_optima/xgb_daegu_smote.pkl')\n",
+    "xgb_daejeon= joblib.load('../save_model/XGB_optima/xgb_daejeon_smote.pkl')\n",
+    "xgb_gwangju= joblib.load('../save_model/XGB_optima/xgb_gwangju_smote.pkl')\n",
+    "\n",
+    "lgb_seoul_1= lgb_seoul[0]\n",
+    "lgb_seoul_2= lgb_seoul[1]\n",
+    "lgb_seoul_3= lgb_seoul[2]\n",
+    "\n",
+    "lgb_busan_1= lgb_busan[0]\n",
+    "lgb_busan_2= lgb_busan[1]\n",
+    "lgb_busan_3= lgb_busan[2]\n",
+    "\n",
+    "lgb_incheon_1= lgb_incheon[0]\n",
+    "lgb_incheon_2= lgb_incheon[1]\n",
+    "lgb_incheon_3= lgb_incheon[2]\n",
+    "\n",
+    "lgb_daegu_1= lgb_daegu[0]\n",
+    "lgb_daegu_2= lgb_daegu[1]\n",
+    "lgb_daegu_3= lgb_daegu[2]\n",
+    "\n",
+    "lgb_daejeon_1= lgb_daejeon[0]\n",
+    "lgb_daejeon_2= lgb_daejeon[1]\n",
+    "lgb_daejeon_3= lgb_daejeon[2]\n",
+    "\n",
+    "lgb_gwangju_1= lgb_gwangju[0]\n",
+    "lgb_gwangju_2= lgb_gwangju[1]\n",
+    "lgb_gwangju_3= lgb_gwangju[2]\n",
+    "\n",
+    "\n",
+    "xgb_seoul_1= xgb_seoul[0]\n",
+    "xgb_seoul_2= xgb_seoul[1]\n",
+    "xgb_seoul_3= xgb_seoul[2]\n",
+    "\n",
+    "xgb_busan_1= xgb_busan[0]\n",
+    "xgb_busan_2= xgb_busan[1]\n",
+    "xgb_busan_3= xgb_busan[2]\n",
+    "\n",
+    "xgb_incheon_1= xgb_incheon[0]\n",
+    "xgb_incheon_2= xgb_incheon[1]\n",
+    "xgb_incheon_3= xgb_incheon[2]\n",
+    "\n",
+    "xgb_daegu_1= xgb_daegu[0]\n",
+    "xgb_daegu_2= xgb_daegu[1]\n",
+    "xgb_daegu_3= xgb_daegu[2]\n",
+    "\n",
+    "xgb_daejeon_1= xgb_daejeon[0]\n",
+    "xgb_daejeon_2= xgb_daejeon[1]\n",
+    "xgb_daejeon_3= xgb_daejeon[2]\n",
+    "\n",
+    "xgb_gwangju_1= xgb_gwangju[0]\n",
+    "xgb_gwangju_2= xgb_gwangju[1]\n",
+    "xgb_gwangju_3= xgb_gwangju[2]\n"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## **Soft Voting**"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## **서울**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "voting = []\n",
+    "mcc = []\n",
+    "accuracy = []\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "Index(['temp_C', 'precip_mm', 'wind_speed', 'wind_dir', 'hm', 'vap_pressure',\n",
+       "       'dewpoint_C', 'loc_pressure', 'sea_pressure', 'solarRad', 'snow_cm',\n",
+       "       'cloudcover', 'lm_cloudcover', 'low_cloudbase', 'groundtemp', 'O3',\n",
+       "       'NO2', 'PM10', 'PM25', 'month', 'hour', 'ground_temp - temp_C',\n",
+       "       'hour_sin', 'hour_cos', 'month_sin', 'month_cos', 'multi_class'],\n",
+       "      dtype='object')"
+      ]
+     },
+     "execution_count": 10,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "df_seoul_test.columns"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "CSI score of soft(test) : 0.3248062015503624\n"
+     ]
+    }
+   ],
+   "source": [
+    "from scipy.stats import mode  # 최다 득표수 계산을 위한 mode 함수\n",
+    "\n",
+    "probas = []\n",
+    "\n",
+    "# 1 Fold\n",
+    "test_preds  = get_proba('seoul', 'deepgbm', 'ctgan20000', 1)\n",
+    "probas.append(test_preds)\n",
+    "test_preds  = get_proba('seoul', 'resnet_like', 'smote', 1)\n",
+    "probas.append(test_preds)\n",
+    "# probas.append(xgb_seoul_1.predict_proba(df_seoul_test.iloc[:,:-1]))\n",
+    "\n",
+    "# 2 Fold\n",
+    "test_preds  = get_proba('seoul', 'deepgbm', 'ctgan20000', 2)\n",
+    "probas.append(test_preds)\n",
+    "test_preds  = get_proba('seoul', 'resnet_like', 'smote', 2)\n",
+    "probas.append(test_preds)\n",
+    "# probas.append(xgb_seoul_2.predict_proba(df_seoul_test.iloc[:,:-1]))\n",
+    "\n",
+    "# 3 Fold\n",
+    "test_preds  = get_proba('seoul', 'deepgbm', 'ctgan20000', 3)\n",
+    "probas.append(test_preds)\n",
+    "test_preds  = get_proba('seoul', 'resnet_like', 'smote', 3)\n",
+    "probas.append(test_preds)\n",
+    "# probas.append(xgb_seoul_3.predict_proba(df_seoul_test.iloc[:,:-1]))\n",
+    "\n",
+    "voting.append(calculate_csi(df_seoul_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "mcc.append(multiclass_mcc(df_seoul_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "accuracy.append(accuracy_score(df_seoul_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "\n",
+    "\n",
+    "print(\"CSI score of soft(test) :\", calculate_csi(df_seoul_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "array([[   2,   11,    0],\n",
+       "       [   6,  417,   58],\n",
+       "       [   7,  789, 7470]])"
+      ]
+     },
+     "execution_count": 12,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "confusion_matrix(df_seoul_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## **부산**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "CSI score of soft(test) : 0.46608315098458075\n"
+     ]
+    }
+   ],
+   "source": [
+    "from scipy.stats import mode  # 최다 득표수 계산을 위한 mode 함수\n",
+    "\n",
+    "probas = []\n",
+    "\n",
+    "# 1 Fold\n",
+    "test_preds  = get_proba('busan', 'deepgbm', 'pure', 1)\n",
+    "probas.append(test_preds)\n",
+    "test_preds  = get_proba('busan', 'resnet_like', 'ctgan10000', 1)\n",
+    "probas.append(test_preds)\n",
+    "\n",
+    "\n",
+    "# 2 Fold\n",
+    "test_preds  = get_proba('busan', 'deepgbm', 'pure', 2)\n",
+    "probas.append(test_preds)\n",
+    "test_preds  = get_proba('busan', 'resnet_like', 'ctgan10000', 2)\n",
+    "probas.append(test_preds)\n",
+    "\n",
+    "\n",
+    "# 3 Fold\n",
+    "test_preds  = get_proba('busan', 'deepgbm', 'pure', 3)\n",
+    "probas.append(test_preds)\n",
+    "test_preds  = get_proba('busan', 'resnet_like', 'ctgan10000', 3)\n",
+    "probas.append(test_preds)\n",
+    "\n",
+    "\n",
+    "voting.append(calculate_csi(df_busan_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "mcc.append(multiclass_mcc(df_busan_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "accuracy.append(accuracy_score(df_busan_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "\n",
+    "print(\"CSI score of soft(test) :\", calculate_csi(df_busan_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "array([[  11,   13,    0],\n",
+       "       [  11,  202,   68],\n",
+       "       [   2,  150, 8303]])"
+      ]
+     },
+     "execution_count": 14,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "confusion_matrix(df_busan_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## **인천**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 15,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "CSI score of hard(test) : 0.572269457161506\n"
+     ]
+    }
+   ],
+   "source": [
+    "from scipy.stats import mode  # 최다 득표수 계산을 위한 mode 함수\n",
+    "\n",
+    "\n",
+    "# 1 Fold\n",
+    "probas = []\n",
+    "test_preds = get_proba('incheon', 'deepgbm', 'pure', 1)\n",
+    "probas.append(test_preds)\n",
+    "test_preds = get_proba('incheon', 'resnet_like', 'smote', 1)\n",
+    "probas.append(test_preds)\n",
+    "test_preds = get_proba('incheon', 'ft_transformer', 'pure', 1)\n",
+    "probas.append(test_preds)\n",
+    "\n",
+    "probas.append(lgb_incheon_1.predict_proba(df_incheon_test.iloc[:,:-1]))\n",
+    "probas.append(xgb_incheon_1.predict_proba(df_incheon_test.iloc[:,:-1]))\n",
+    "\n",
+    "# 2 Fold\n",
+    "test_preds = get_proba('incheon', 'deepgbm', 'pure', 2)\n",
+    "probas.append(test_preds)\n",
+    "test_preds = get_proba('incheon', 'resnet_like', 'smote', 2)\n",
+    "probas.append(test_preds)\n",
+    "test_preds = get_proba('incheon', 'ft_transformer', 'pure', 2)\n",
+    "probas.append(test_preds)\n",
+    "\n",
+    "probas.append(lgb_incheon_2.predict_proba(df_incheon_test.iloc[:,:-1]))\n",
+    "probas.append(xgb_incheon_2.predict_proba(df_incheon_test.iloc[:,:-1]))\n",
+    "\n",
+    "# 3 Fold\n",
+    "test_preds = get_proba('incheon', 'deepgbm', 'pure', 3)\n",
+    "probas.append(test_preds)\n",
+    "test_preds = get_proba('incheon', 'resnet_like', 'smote', 3)\n",
+    "probas.append(test_preds)\n",
+    "test_preds = get_proba('incheon', 'ft_transformer', 'pure', 3)\n",
+    "probas.append(test_preds)\n",
+    "\n",
+    "probas.append(lgb_incheon_3.predict_proba(df_incheon_test.iloc[:,:-1]))\n",
+    "probas.append(xgb_incheon_3.predict_proba(df_incheon_test.iloc[:,:-1]))\n",
+    "\n",
+    "\n",
+    "\n",
+    "voting.append(calculate_csi(df_incheon_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "mcc.append(multiclass_mcc(df_incheon_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "accuracy.append(accuracy_score(df_incheon_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "\n",
+    "\n",
+    "print(\"CSI score of hard(test) :\", calculate_csi(df_incheon_test.iloc[:,-1],mode(np.argmax(probas, axis=2), axis=0).mode[0]))\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 16,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "array([[  87,   74,   21],\n",
+       "       [  22,  788,  395],\n",
+       "       [   2,  140, 7231]])"
+      ]
+     },
+     "execution_count": 16,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "confusion_matrix(df_incheon_test.iloc[:,-1],mode(np.argmax(probas, axis=2), axis=0).mode[0])"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## **대구**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 17,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "CSI score of soft(test) : 0.2852112676055334\n"
+     ]
+    }
+   ],
+   "source": [
+    "from scipy.stats import mode  # 최다 득표수 계산을 위한 mode 함수\n",
+    "\n",
+    "probas= []\n",
+    "\n",
+    "# 1 Fold\n",
+    "test_preds = get_proba('daegu', 'deepgbm', 'smote', 1)\n",
+    "probas.append(test_preds)\n",
+    "test_preds = get_proba('daegu', 'ft_transformer', 'pure', 1)\n",
+    "probas.append(test_preds)\n",
+    "\n",
+    "# 2 Fold\n",
+    "test_preds = get_proba('daegu', 'deepgbm', 'smote', 2)\n",
+    "probas.append(test_preds)\n",
+    "test_preds = get_proba('daegu', 'ft_transformer', 'pure', 2)\n",
+    "probas.append(test_preds)\n",
+    "\n",
+    "# 3 Fold\n",
+    "test_preds = get_proba('daegu', 'deepgbm', 'smote', 3)\n",
+    "probas.append(test_preds)\n",
+    "test_preds = get_proba('daegu', 'ft_transformer', 'pure', 3)\n",
+    "probas.append(test_preds)\n",
+    "\n",
+    "voting.append(calculate_csi(df_daegu_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "mcc.append(multiclass_mcc(df_daegu_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "accuracy.append(accuracy_score(df_daegu_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "\n",
+    "print(\"CSI score of soft(test) :\", calculate_csi(df_daegu_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 18,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "array([[   1,    0,    0],\n",
+       "       [   1,   80,   47],\n",
+       "       [   2,  153, 8476]])"
+      ]
+     },
+     "execution_count": 18,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "confusion_matrix(df_daegu_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## **대전**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 19,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "CSI score of soft(test) : 0.31884057971011603\n"
+     ]
+    }
+   ],
+   "source": [
+    "from scipy.stats import mode  # 최다 득표수 계산을 위한 mode 함수\n",
+    "\n",
+    "probas = []\n",
+    "\n",
+    "# 1 Fold\n",
+    "test_preds = get_proba('daejeon', 'deepgbm', 'pure', 1)\n",
+    "probas.append(test_preds)\n",
+    "test_preds = get_proba('daejeon', 'ft_transformer', 'pure', 1)\n",
+    "probas.append(test_preds)\n",
+    "\n",
+    "\n",
+    "# 2 Fold\n",
+    "test_preds = get_proba('daejeon', 'deepgbm', 'pure', 2)\n",
+    "probas.append(test_preds)\n",
+    "test_preds = get_proba('daejeon', 'ft_transformer', 'pure', 2)\n",
+    "probas.append(test_preds)\n",
+    "\n",
+    "\n",
+    "# 3 Fold\n",
+    "test_preds = get_proba('daejeon', 'deepgbm', 'pure', 3)\n",
+    "probas.append(test_preds)\n",
+    "test_preds = get_proba('daejeon', 'ft_transformer', 'pure', 3)\n",
+    "probas.append(test_preds)\n",
+    "\n",
+    "voting.append(calculate_csi(df_daejeon_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "mcc.append(multiclass_mcc(df_daejeon_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "accuracy.append(accuracy_score(df_daejeon_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "\n",
+    "\n",
+    "print(\"CSI score of soft(test) :\", calculate_csi(df_daejeon_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 20,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "array([[  15,   23,   15],\n",
+       "       [  10,  337,  271],\n",
+       "       [   0,  433, 7656]])"
+      ]
+     },
+     "execution_count": 20,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "confusion_matrix(df_daejeon_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1))"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## **광주**"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 21,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "CSI score of soft(test) : 0.4759725400457121\n"
+     ]
+    }
+   ],
+   "source": [
+    "from scipy.stats import mode  # 최다 득표수 계산을 위한 mode 함수\n",
+    "\n",
+    "probas = []\n",
+    "\n",
+    "# 1 Fold\n",
+    "test_preds = get_proba('gwangju', 'deepgbm', 'pure', 1)\n",
+    "probas.append(test_preds)\n",
+    "test_preds = get_proba('gwangju', 'ft_transformer', 'pure', 1)\n",
+    "probas.append(test_preds)\n",
+    "\n",
+    "\n",
+    "# 2 Fold\n",
+    "test_preds = get_proba('gwangju', 'deepgbm', 'pure', 2)\n",
+    "probas.append(test_preds)\n",
+    "test_preds = get_proba('gwangju', 'ft_transformer', 'pure', 2)\n",
+    "probas.append(test_preds)\n",
+    "\n",
+    "\n",
+    "# 3 Fold\n",
+    "test_preds = get_proba('gwangju', 'deepgbm', 'pure', 3)\n",
+    "probas.append(test_preds)\n",
+    "test_preds = get_proba('gwangju', 'ft_transformer', 'pure', 3)\n",
+    "probas.append(test_preds)\n",
+    "\n",
+    "voting.append(calculate_csi(df_gwangju_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "mcc.append(multiclass_mcc(df_gwangju_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "accuracy.append(accuracy_score(df_gwangju_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))\n",
+    "\n",
+    "\n",
+    "print(\"CSI score of soft(test) :\", calculate_csi(df_gwangju_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1)))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 22,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "array([[  10,   12,    8],\n",
+       "       [   2,  406,  235],\n",
+       "       [   0,  201, 7886]])"
+      ]
+     },
+     "execution_count": 22,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "confusion_matrix(df_gwangju_test.iloc[:,-1], np.argmax(np.mean(probas, axis=0), axis=1))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 23,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "image/png": "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",
+      "text/plain": [
+       "<Figure size 1000x500 with 1 Axes>"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    }
+   ],
+   "source": [
+    "import matplotlib.pyplot as plt\n",
+    "import seaborn as sns\n",
+    "\n",
+    "plt.figure(figsize=(10,5))\n",
+    "sns.barplot(x=['Seoul', 'Busan', 'Incheon', 'Daegu', 'Daejeon', 'Gwangju'], y=voting)\n",
+    "plt.title('CSI Score of Voting Model')\n",
+    "plt.ylabel('CSI')\n",
+    "\n",
+    "plt.show()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 24,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "0.4078715882283252"
+      ]
+     },
+     "execution_count": 24,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "np.mean(voting)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 25,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "[0.3248062015503624,\n",
+       " 0.46608315098458075,\n",
+       " 0.5763157894736463,\n",
+       " 0.2852112676055334,\n",
+       " 0.31884057971011603,\n",
+       " 0.4759725400457121]"
+      ]
+     },
+     "execution_count": 25,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "voting"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 26,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "[0.5106142349456536,\n",
+       " 0.640202275543952,\n",
+       " 0.709448778435959,\n",
+       " 0.45579515959653394,\n",
+       " 0.453960121993875,\n",
+       " 0.6218724605270242]"
+      ]
+     },
+     "execution_count": 26,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "mcc"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 27,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "[0.9005707762557078,\n",
+       " 0.9721461187214612,\n",
+       " 0.9264840182648402,\n",
+       " 0.9768264840182649,\n",
+       " 0.9141552511415525,\n",
+       " 0.9477168949771689]"
+      ]
+     },
+     "execution_count": 27,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "accuracy"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.8.10"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

Analysis_code/ft_transformer.py

@@ -0,0 +1,56 @@
+import torch
+import torch.nn as nn
+
+# FT-Transformer Implementation
+class FTTransformer(nn.Module):
+    def __init__(self, num_features, cat_cardinalities, num_classes, d_token=192, n_blocks=6, attention_dropout=0.2, ffn_dropout=0.2):
+        super(FTTransformer, self).__init__()
+
+        self.num_classes = num_classes  # 클래스 개수 저장
+
+        # Embedding layers for categorical features
+        self.cat_embeddings = nn.ModuleList([
+            nn.Embedding(num_categories, d_token) for num_categories in cat_cardinalities
+        ])
+
+        # Linear layer for numerical features
+        self.num_linear = nn.Linear(num_features, d_token)
+
+        # Transformer blocks
+        self.transformer_blocks = nn.ModuleList([
+            nn.TransformerEncoderLayer(
+                d_model=d_token,
+                nhead=8,
+                dim_feedforward=4 * d_token,
+                dropout=attention_dropout,
+                activation='gelu'
+            ) for _ in range(n_blocks)
+        ])
+
+        self.ffn_dropout = nn.Dropout(ffn_dropout)
+        if num_classes == 2:
+            self.output_layer = nn.Linear(d_token, 1)  # Binary classification
+        elif num_classes > 2:
+            self.output_layer = nn.Linear(d_token, num_classes)  # Multi classification
+
+    def forward(self, x_num, x_cat):
+        # Numerical feature embedding
+        x_num = self.num_linear(x_num)
+
+        # Categorical feature embedding
+        x_cat = [embed(x_cat[:, i]) for i, embed in enumerate(self.cat_embeddings)]
+        x_cat = torch.stack(x_cat, dim=1)
+
+        # Combine numerical and categorical embeddings
+        x = x_num.unsqueeze(1) + x_cat
+
+        # Pass through transformer blocks
+        for block in self.transformer_blocks:
+            x = block(x)
+
+        # Pooling and output
+        x = x.mean(dim=1)
+        x = self.ffn_dropout(x)
+        x = self.output_layer(x)
+
+        return x

Analysis_code/make_oversample_data/gan_sample_10000_1.py

@@ -0,0 +1,181 @@
+import pandas as pd
+import numpy as np
+
+from imblearn.over_sampling import SMOTENC
+import optuna
+from ctgan import CTGAN
+import torch
+import warnings
+
+# 지역별 데이터 파일 경로
+regions = ['incheon', 'seoul','busan', 'daegu', 'daejeon', 'gwangju']
+file_paths = [f'../../data/data_for_modeling/{region}_train.csv' for region in regions]
+output_paths = [f'../../data/data_oversampled/ctgan10000/ctgan10000_1_{region}.csv' for region in regions]
+
+# GPU 사용 설정
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+print(f"Using device: {device}")
+
+# 경고 무시
+warnings.filterwarnings("ignore", category=UserWarning, module="optuna.distributions")
+
+# 지역별 처리
+for file_path, output_path in zip(file_paths, output_paths):
+    # 데이터 로드
+    data = pd.read_csv(file_path, index_col=0)
+    data= data.loc[data['year'].isin([2018,2019]),:]
+    data['cloudcover'] = data['cloudcover'].astype('int')
+    data['lm_cloudcover'] = data['lm_cloudcover'].astype('int')
+    X = data.drop(columns=['multi_class', 'binary_class'])
+    y = data['multi_class']
+
+    # 불필요한 열 제거
+    X.drop(columns=['ground_temp - temp_C', 'hour_sin', 'hour_cos', 'month_sin', 'month_cos'], inplace=True)
+
+    # SMOTENC에서 사용할 범주형 변수 열 번호 설정
+    categorical_features_indices = [i for i, dtype in enumerate(X.dtypes) if dtype != 'float64']
+
+    # sampling_strategy 설정
+    count_class_0 = (y == 0).sum()
+    count_class_1 = (y == 1).sum()
+    count_class_2 = (y == 2).sum()
+    sampling_strategy = {
+        0: 500 if count_class_0 <= 500 else 1000,
+        1: int(np.ceil(count_class_1 / 100) * 100),  # 백의 자리로 올림
+        2: count_class_2
+    }
+
+    # SMOTENC 적용
+    smotenc = SMOTENC(categorical_features=categorical_features_indices, sampling_strategy=sampling_strategy, random_state=42)
+    X_resampled, y_resampled = smotenc.fit_resample(X, y)
+
+    # Resampled 데이터 생성
+    lerp_data = X_resampled.copy()
+    lerp_data['multi_class'] = y_resampled
+
+    # CTGAN에서 사용할 범주형 변수 열 이름 설정
+    categorical_features = [
+        col for col, dtype in zip(lerp_data.columns, lerp_data.dtypes) if dtype != 'float64'
+    ]
+
+    # Optuna 목적 함수 정의
+    def objective(trial):
+        # 하이퍼파라미터 탐색 범위 설정
+        embedding_dim = trial.suggest_int("embedding_dim", 64, 128)
+        generator_dim = trial.suggest_categorical("generator_dim", [(64, 64), (128, 128)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(64, 64), (128, 128)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [64, 128, 256])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 3)
+
+        # CTGAN 모델 생성
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        # 범주 0 데이터 필터링
+        data_0 = lerp_data[lerp_data['multi_class'] == 0]
+
+        # 모델 학습
+        ctgan.fit(data_0, discrete_columns=categorical_features)
+
+        # 샘플 생성
+        generated_data = ctgan.sample(len(data_0) * 2)
+
+        # 평가: 샘플의 연속형 변수 분포 비교
+        real_visi = data_0['visi']
+        generated_visi = generated_data['visi']
+
+        # 분포 간 차이(MSE) 계산
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    # Optuna로 최적화 수행
+    study = optuna.create_study(direction="maximize")
+    study.optimize(objective, n_trials=50)
+
+    # 최적 하이퍼파라미터 출력
+    best_params = study.best_params
+
+    # 최적 하이퍼파라미터로 CTGAN 학습 및 샘플 생성
+    ctgan = CTGAN(
+        embedding_dim=best_params["embedding_dim"],
+        generator_dim=best_params["generator_dim"],
+        discriminator_dim=best_params["discriminator_dim"],
+        batch_size=best_params["batch_size"],
+        discriminator_steps=best_params["discriminator_steps"],
+        pac=best_params["pac"]
+    )
+
+    # 범주 0 데이터로 최종 학습
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 0], discrete_columns=categorical_features)
+    generated_0 = ctgan.sample(10000)
+
+    # 범주 1 데이터 최적화 및 생성
+    def objective_class1(trial):
+        embedding_dim = trial.suggest_int("embedding_dim", 128, 512)
+        generator_dim = trial.suggest_categorical("generator_dim", [(128, 128), (256, 256)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(128, 128), (256, 256)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [256, 512, 1024])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 5)
+
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        data_1 = lerp_data[lerp_data['multi_class'] == 1]
+        ctgan.fit(data_1, discrete_columns=categorical_features)
+        generated_data = ctgan.sample(len(data_1) * 2)
+
+        real_visi = data_1['visi']
+        generated_visi = generated_data['visi']
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    study_class1 = optuna.create_study(direction="maximize")
+    study_class1.optimize(objective_class1, n_trials=30)
+
+    best_params_class1 = study_class1.best_params
+    ctgan = CTGAN(
+        embedding_dim=best_params_class1["embedding_dim"],
+        generator_dim=best_params_class1["generator_dim"],
+        discriminator_dim=best_params_class1["discriminator_dim"],
+        batch_size=best_params_class1["batch_size"],
+        discriminator_steps=best_params_class1["discriminator_steps"],
+        pac=best_params_class1["pac"]
+    )
+
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 1], discrete_columns=categorical_features)
+    generated_1 = ctgan.sample(10000 - int(np.ceil(count_class_1 / 100) * 100))
+
+    # 데이터 병합 및 저장
+    well_generated0 = generated_0[(generated_0['visi'] >= 0) & (generated_0['visi'] < 100)]
+    well_generated1 = generated_1[(generated_1['visi'] >= 100) & (generated_1['visi'] < 500)]
+    smote_gan_data = pd.concat([lerp_data, well_generated0, well_generated1], axis=0)
+    # 제거변수 복구
+    smote_gan_data['binary_class'] = smote_gan_data['multi_class'].apply(lambda x: 0 if x == 2 else 1)
+    smote_gan_data['hour_sin'] = np.sin(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['hour_cos'] = np.cos(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['month_sin'] = np.sin(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['month_cos'] = np.cos(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['ground_temp - temp_C'] = smote_gan_data['groundtemp'] - smote_gan_data['temp_C']
+
+    filtered_data = smote_gan_data[smote_gan_data['multi_class'] != 2]
+    original_class2 = data[data['multi_class'] == 2]
+    final_data = pd.concat([filtered_data, original_class2], axis=0)
+    final_data.reset_index(drop=True, inplace=True)
+
+    # 결과 저장
+    final_data.to_csv(output_path, index = False)
+    print(len(final_data[final_data['multi_class']==0]),'|',len(final_data[final_data['multi_class']==1]),'|',len(final_data[final_data['multi_class']==2]))

Analysis_code/make_oversample_data/gan_sample_10000_2.py

@@ -0,0 +1,182 @@
+import pandas as pd
+import numpy as np
+import os
+os.environ["CUDA_VISIBLE_DEVICES"] = "1"
+from imblearn.over_sampling import SMOTENC
+import optuna
+from ctgan import CTGAN
+import torch
+import warnings
+
+# 지역별 데이터 파일 경로
+regions = ['incheon', 'seoul','busan', 'daegu', 'daejeon', 'gwangju']
+file_paths = [f'../../data/data_for_modeling/{region}_train.csv' for region in regions]
+output_paths = [f'../../data/data_oversampled/ctgan10000/ctgan10000_2_{region}.csv' for region in regions]
+
+# GPU 사용 설정
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+print(f"Using device: {device}")
+
+# 경고 무시
+warnings.filterwarnings("ignore", category=UserWarning, module="optuna.distributions")
+
+# 지역별 처리
+for file_path, output_path in zip(file_paths, output_paths):
+    # 데이터 로드
+    data = pd.read_csv(file_path, index_col=0)
+    data= data.loc[data['year'].isin([2018,2020]),:]
+    data['cloudcover'] = data['cloudcover'].astype('int')
+    data['lm_cloudcover'] = data['lm_cloudcover'].astype('int')
+    X = data.drop(columns=['multi_class', 'binary_class'])
+    y = data['multi_class']
+
+    # 불필요한 열 제거
+    X.drop(columns=['ground_temp - temp_C', 'hour_sin', 'hour_cos', 'month_sin', 'month_cos'], inplace=True)
+
+    # SMOTENC에서 사용할 범주형 변수 열 번호 설정
+    categorical_features_indices = [i for i, dtype in enumerate(X.dtypes) if dtype != 'float64']
+
+    # sampling_strategy 설정
+    count_class_0 = (y == 0).sum()
+    count_class_1 = (y == 1).sum()
+    count_class_2 = (y == 2).sum()
+    sampling_strategy = {
+        0: 500 if count_class_0 <= 500 else 1000,
+        1: int(np.ceil(count_class_1 / 100) * 100),  # 백의 자리로 올림
+        2: count_class_2
+    }
+
+    # SMOTENC 적용
+    smotenc = SMOTENC(categorical_features=categorical_features_indices, sampling_strategy=sampling_strategy, random_state=42)
+    X_resampled, y_resampled = smotenc.fit_resample(X, y)
+
+    # Resampled 데이터 생성
+    lerp_data = X_resampled.copy()
+    lerp_data['multi_class'] = y_resampled
+
+    # CTGAN에서 사용할 범주형 변수 열 이름 설정
+    categorical_features = [
+        col for col, dtype in zip(lerp_data.columns, lerp_data.dtypes) if dtype != 'float64'
+    ]
+
+    # Optuna 목적 함수 정의
+    def objective(trial):
+        # 하이퍼파라미터 탐색 범위 설정
+        embedding_dim = trial.suggest_int("embedding_dim", 64, 128)
+        generator_dim = trial.suggest_categorical("generator_dim", [(64, 64), (128, 128)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(64, 64), (128, 128)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [64, 128, 256])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 3)
+
+        # CTGAN 모델 생성
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        # 범주 0 데이터 필터링
+        data_0 = lerp_data[lerp_data['multi_class'] == 0]
+
+        # 모델 학습
+        ctgan.fit(data_0, discrete_columns=categorical_features)
+
+        # 샘플 생성
+        generated_data = ctgan.sample(len(data_0) * 2)
+
+        # 평가: 샘플의 연속형 변수 분포 비교
+        real_visi = data_0['visi']
+        generated_visi = generated_data['visi']
+
+        # 분포 간 차이(MSE) 계산
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    # Optuna로 최적화 수행
+    study = optuna.create_study(direction="maximize")
+    study.optimize(objective, n_trials=50)
+
+    # 최적 하이퍼파라미터 출력
+    best_params = study.best_params
+
+    # 최적 하이퍼파라미터로 CTGAN 학습 및 샘플 생성
+    ctgan = CTGAN(
+        embedding_dim=best_params["embedding_dim"],
+        generator_dim=best_params["generator_dim"],
+        discriminator_dim=best_params["discriminator_dim"],
+        batch_size=best_params["batch_size"],
+        discriminator_steps=best_params["discriminator_steps"],
+        pac=best_params["pac"]
+    )
+
+    # 범주 0 데이터로 최종 학습
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 0], discrete_columns=categorical_features)
+    generated_0 = ctgan.sample(10000)
+
+    # 범주 1 데이터 최적화 및 생성
+    def objective_class1(trial):
+        embedding_dim = trial.suggest_int("embedding_dim", 128, 512)
+        generator_dim = trial.suggest_categorical("generator_dim", [(128, 128), (256, 256)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(128, 128), (256, 256)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [256, 512, 1024])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 5)
+
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        data_1 = lerp_data[lerp_data['multi_class'] == 1]
+        ctgan.fit(data_1, discrete_columns=categorical_features)
+        generated_data = ctgan.sample(len(data_1) * 2)
+
+        real_visi = data_1['visi']
+        generated_visi = generated_data['visi']
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    study_class1 = optuna.create_study(direction="maximize")
+    study_class1.optimize(objective_class1, n_trials=30)
+
+    best_params_class1 = study_class1.best_params
+    ctgan = CTGAN(
+        embedding_dim=best_params_class1["embedding_dim"],
+        generator_dim=best_params_class1["generator_dim"],
+        discriminator_dim=best_params_class1["discriminator_dim"],
+        batch_size=best_params_class1["batch_size"],
+        discriminator_steps=best_params_class1["discriminator_steps"],
+        pac=best_params_class1["pac"]
+    )
+
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 1], discrete_columns=categorical_features)
+    generated_1 = ctgan.sample(10000 - int(np.ceil(count_class_1 / 100) * 100))
+
+    # 데이터 병합 및 저장
+    well_generated0 = generated_0[(generated_0['visi'] >= 0) & (generated_0['visi'] < 100)]
+    well_generated1 = generated_1[(generated_1['visi'] >= 100) & (generated_1['visi'] < 500)]
+    smote_gan_data = pd.concat([lerp_data, well_generated0, well_generated1], axis=0)
+    # 제거변수 복구
+    smote_gan_data['binary_class'] = smote_gan_data['multi_class'].apply(lambda x: 0 if x == 2 else 1)
+    smote_gan_data['hour_sin'] = np.sin(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['hour_cos'] = np.cos(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['month_sin'] = np.sin(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['month_cos'] = np.cos(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['ground_temp - temp_C'] = smote_gan_data['groundtemp'] - smote_gan_data['temp_C']
+
+    filtered_data = smote_gan_data[smote_gan_data['multi_class'] != 2]
+    original_class2 = data[data['multi_class'] == 2]
+    final_data = pd.concat([filtered_data, original_class2], axis=0)
+    final_data.reset_index(drop=True, inplace=True)
+
+    # 결과 저장
+    final_data.to_csv(output_path, index = False)
+    print(len(final_data[final_data['multi_class']==0]),'|',len(final_data[final_data['multi_class']==1]),'|',len(final_data[final_data['multi_class']==2]))

Analysis_code/make_oversample_data/gan_sample_10000_3.py

@@ -0,0 +1,182 @@
+import pandas as pd
+import numpy as np
+import os
+os.environ["CUDA_VISIBLE_DEVICES"] = "0"
+from imblearn.over_sampling import SMOTENC
+import optuna
+from ctgan import CTGAN
+import torch
+import warnings
+
+# 지역별 데이터 파일 경로
+regions = ['incheon', 'seoul','busan', 'daegu', 'daejeon', 'gwangju']
+file_paths = [f'../../data/data_for_modeling/{region}_train.csv' for region in regions]
+output_paths = [f'../../data/data_oversampled/ctgan10000/ctgan10000_3_{region}.csv' for region in regions]
+
+# GPU 사용 설정
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+print(f"Using device: {device}")
+
+# 경고 무시
+warnings.filterwarnings("ignore", category=UserWarning, module="optuna.distributions")
+
+# 지역별 처리
+for file_path, output_path in zip(file_paths, output_paths):
+    # 데이터 로드
+    data = pd.read_csv(file_path, index_col=0)
+    data= data.loc[data['year'].isin([2019,2020]),:]
+    data['cloudcover'] = data['cloudcover'].astype('int')
+    data['lm_cloudcover'] = data['lm_cloudcover'].astype('int')
+    X = data.drop(columns=['multi_class', 'binary_class'])
+    y = data['multi_class']
+
+    # 불필요한 열 제거
+    X.drop(columns=['ground_temp - temp_C', 'hour_sin', 'hour_cos', 'month_sin', 'month_cos'], inplace=True)
+
+    # SMOTENC에서 사용할 범주형 변수 열 번호 설정
+    categorical_features_indices = [i for i, dtype in enumerate(X.dtypes) if dtype != 'float64']
+
+    # sampling_strategy 설정
+    count_class_0 = (y == 0).sum()
+    count_class_1 = (y == 1).sum()
+    count_class_2 = (y == 2).sum()
+    sampling_strategy = {
+        0: 500 if count_class_0 <= 500 else 1000,
+        1: int(np.ceil(count_class_1 / 100) * 100),  # 백의 자리로 올림
+        2: count_class_2
+    }
+
+    # SMOTENC 적용
+    smotenc = SMOTENC(categorical_features=categorical_features_indices, sampling_strategy=sampling_strategy, random_state=42)
+    X_resampled, y_resampled = smotenc.fit_resample(X, y)
+
+    # Resampled 데이터 생성
+    lerp_data = X_resampled.copy()
+    lerp_data['multi_class'] = y_resampled
+
+    # CTGAN에서 사용할 범주형 변수 열 이름 설정
+    categorical_features = [
+        col for col, dtype in zip(lerp_data.columns, lerp_data.dtypes) if dtype != 'float64'
+    ]
+
+    # Optuna 목적 함수 정의
+    def objective(trial):
+        # 하이퍼파라미터 탐색 범위 설정
+        embedding_dim = trial.suggest_int("embedding_dim", 64, 128)
+        generator_dim = trial.suggest_categorical("generator_dim", [(64, 64), (128, 128)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(64, 64), (128, 128)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [64, 128, 256])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 3)
+
+        # CTGAN 모델 생성
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        # 범주 0 데이터 필터링
+        data_0 = lerp_data[lerp_data['multi_class'] == 0]
+
+        # 모델 학습
+        ctgan.fit(data_0, discrete_columns=categorical_features)
+
+        # 샘플 생성
+        generated_data = ctgan.sample(len(data_0) * 2)
+
+        # 평가: 샘플의 연속형 변수 분포 비교
+        real_visi = data_0['visi']
+        generated_visi = generated_data['visi']
+
+        # 분포 간 차이(MSE) 계산
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    # Optuna로 최적화 수행
+    study = optuna.create_study(direction="maximize")
+    study.optimize(objective, n_trials=50)
+
+    # 최적 하이퍼파라미터 출력
+    best_params = study.best_params
+
+    # 최적 하이퍼파라미터로 CTGAN 학습 및 샘플 생성
+    ctgan = CTGAN(
+        embedding_dim=best_params["embedding_dim"],
+        generator_dim=best_params["generator_dim"],
+        discriminator_dim=best_params["discriminator_dim"],
+        batch_size=best_params["batch_size"],
+        discriminator_steps=best_params["discriminator_steps"],
+        pac=best_params["pac"]
+    )
+
+    # 범주 0 데이터로 최종 학습
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 0], discrete_columns=categorical_features)
+    generated_0 = ctgan.sample(10000)
+
+    # 범주 1 데이터 최적화 및 생성
+    def objective_class1(trial):
+        embedding_dim = trial.suggest_int("embedding_dim", 128, 512)
+        generator_dim = trial.suggest_categorical("generator_dim", [(128, 128), (256, 256)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(128, 128), (256, 256)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [256, 512, 1024])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 5)
+
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        data_1 = lerp_data[lerp_data['multi_class'] == 1]
+        ctgan.fit(data_1, discrete_columns=categorical_features)
+        generated_data = ctgan.sample(len(data_1) * 2)
+
+        real_visi = data_1['visi']
+        generated_visi = generated_data['visi']
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    study_class1 = optuna.create_study(direction="maximize")
+    study_class1.optimize(objective_class1, n_trials=30)
+
+    best_params_class1 = study_class1.best_params
+    ctgan = CTGAN(
+        embedding_dim=best_params_class1["embedding_dim"],
+        generator_dim=best_params_class1["generator_dim"],
+        discriminator_dim=best_params_class1["discriminator_dim"],
+        batch_size=best_params_class1["batch_size"],
+        discriminator_steps=best_params_class1["discriminator_steps"],
+        pac=best_params_class1["pac"]
+    )
+
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 1], discrete_columns=categorical_features)
+    generated_1 = ctgan.sample(10000 - int(np.ceil(count_class_1 / 100) * 100))
+
+    # 데이터 병합 및 저장
+    well_generated0 = generated_0[(generated_0['visi'] >= 0) & (generated_0['visi'] < 100)]
+    well_generated1 = generated_1[(generated_1['visi'] >= 100) & (generated_1['visi'] < 500)]
+    smote_gan_data = pd.concat([lerp_data, well_generated0, well_generated1], axis=0)
+    # 제거변수 복구
+    smote_gan_data['binary_class'] = smote_gan_data['multi_class'].apply(lambda x: 0 if x == 2 else 1)
+    smote_gan_data['hour_sin'] = np.sin(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['hour_cos'] = np.cos(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['month_sin'] = np.sin(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['month_cos'] = np.cos(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['ground_temp - temp_C'] = smote_gan_data['groundtemp'] - smote_gan_data['temp_C']
+
+    filtered_data = smote_gan_data[smote_gan_data['multi_class'] != 2]
+    original_class2 = data[data['multi_class'] == 2]
+    final_data = pd.concat([filtered_data, original_class2], axis=0)
+    final_data.reset_index(drop=True, inplace=True)
+
+    # 결과 저장
+    final_data.to_csv(output_path, index = False)
+    print(len(final_data[final_data['multi_class']==0]),'|',len(final_data[final_data['multi_class']==1]),'|',len(final_data[final_data['multi_class']==2]))

Analysis_code/make_oversample_data/gan_sample_20000_1.py

@@ -0,0 +1,183 @@
+import pandas as pd
+import numpy as np
+# import os
+# os.environ["CUDA_VISIBLE_DEVICES"] = "1"
+from imblearn.over_sampling import SMOTENC
+import optuna
+from ctgan import CTGAN
+import torch
+import warnings
+
+# 지역별 데이터 파일 경로
+# regions = ['busan', 'daegu', 'daejeon', 'gwangju', 'incheon', 'seoul']
+regions = ['incheon', 'seoul','busan', 'daegu', 'daejeon', 'gwangju']
+file_paths = [f'../../data/data_for_modeling/{region}_train.csv' for region in regions]
+output_paths = [f'../../data/data_oversampled/ctgan20000/ctgan20000_1_{region}.csv' for region in regions]
+
+# GPU 사용 설정
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+print(f"Using device: {device}")
+
+# 경고 무시
+warnings.filterwarnings("ignore", category=UserWarning, module="optuna.distributions")
+
+# 지역별 처리
+for file_path, output_path in zip(file_paths, output_paths):
+    # 데이터 로드
+    data = pd.read_csv(file_path, index_col=0)
+    data= data.loc[data['year'].isin([2018,2019]),:]
+    data['cloudcover'] = data['cloudcover'].astype('int')
+    data['lm_cloudcover'] = data['lm_cloudcover'].astype('int')
+    X = data.drop(columns=['multi_class', 'binary_class'])
+    y = data['multi_class']
+
+    # 불필요한 열 제거
+    X.drop(columns=['ground_temp - temp_C', 'hour_sin', 'hour_cos', 'month_sin', 'month_cos'], inplace=True)
+
+    # SMOTENC에서 사용할 범주형 변수 열 번호 설정
+    categorical_features_indices = [i for i, dtype in enumerate(X.dtypes) if dtype != 'float64']
+
+    # sampling_strategy 설정
+    count_class_0 = (y == 0).sum()
+    count_class_1 = (y == 1).sum()
+    count_class_2 = (y == 2).sum()
+    sampling_strategy = {
+        0: 500 if count_class_0 <= 500 else 1000,
+        1: int(np.ceil(count_class_1 / 100) * 100),  # 백의 자리로 올림
+        2: count_class_2
+    }
+
+    # SMOTENC 적용
+    smotenc = SMOTENC(categorical_features=categorical_features_indices, sampling_strategy=sampling_strategy, random_state=42)
+    X_resampled, y_resampled = smotenc.fit_resample(X, y)
+
+    # Resampled 데이터 생성
+    lerp_data = X_resampled.copy()
+    lerp_data['multi_class'] = y_resampled
+
+    # CTGAN에서 사용할 범주형 변수 열 이름 설정
+    categorical_features = [
+        col for col, dtype in zip(lerp_data.columns, lerp_data.dtypes) if dtype != 'float64'
+    ]
+
+    # Optuna 목적 함수 정의
+    def objective(trial):
+        # 하이퍼파라미터 탐색 범위 설정
+        embedding_dim = trial.suggest_int("embedding_dim", 64, 128)
+        generator_dim = trial.suggest_categorical("generator_dim", [(64, 64), (128, 128)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(64, 64), (128, 128)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [64, 128, 256])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 3)
+
+        # CTGAN 모델 생성
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        # 범주 0 데이터 필터링
+        data_0 = lerp_data[lerp_data['multi_class'] == 0]
+
+        # 모델 학습
+        ctgan.fit(data_0, discrete_columns=categorical_features)
+
+        # 샘플 생성
+        generated_data = ctgan.sample(len(data_0) * 2)
+
+        # 평가: 샘플의 연속형 변수 분포 비교
+        real_visi = data_0['visi']
+        generated_visi = generated_data['visi']
+
+        # 분포 간 차이(MSE) 계산
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    # Optuna로 최적화 수행
+    study = optuna.create_study(direction="maximize")
+    study.optimize(objective, n_trials=50)
+
+    # 최적 하이퍼파라미터 출력
+    best_params = study.best_params
+
+    # 최적 하이퍼파라미터로 CTGAN 학습 및 샘플 생성
+    ctgan = CTGAN(
+        embedding_dim=best_params["embedding_dim"],
+        generator_dim=best_params["generator_dim"],
+        discriminator_dim=best_params["discriminator_dim"],
+        batch_size=best_params["batch_size"],
+        discriminator_steps=best_params["discriminator_steps"],
+        pac=best_params["pac"]
+    )
+
+    # 범주 0 데이터로 최종 학습
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 0], discrete_columns=categorical_features)
+    generated_0 = ctgan.sample(20000)
+
+    # 범주 1 데이터 최적화 및 생성
+    def objective_class1(trial):
+        embedding_dim = trial.suggest_int("embedding_dim", 128, 512)
+        generator_dim = trial.suggest_categorical("generator_dim", [(128, 128), (256, 256)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(128, 128), (256, 256)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [256, 512, 1024])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 5)
+
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        data_1 = lerp_data[lerp_data['multi_class'] == 1]
+        ctgan.fit(data_1, discrete_columns=categorical_features)
+        generated_data = ctgan.sample(len(data_1) * 2)
+
+        real_visi = data_1['visi']
+        generated_visi = generated_data['visi']
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    study_class1 = optuna.create_study(direction="maximize")
+    study_class1.optimize(objective_class1, n_trials=30)
+
+    best_params_class1 = study_class1.best_params
+    ctgan = CTGAN(
+        embedding_dim=best_params_class1["embedding_dim"],
+        generator_dim=best_params_class1["generator_dim"],
+        discriminator_dim=best_params_class1["discriminator_dim"],
+        batch_size=best_params_class1["batch_size"],
+        discriminator_steps=best_params_class1["discriminator_steps"],
+        pac=best_params_class1["pac"]
+    )
+
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 1], discrete_columns=categorical_features)
+    generated_1 = ctgan.sample(20000 - int(np.ceil(count_class_1 / 100) * 100))
+
+    # 데이터 병합 및 저장
+    well_generated0 = generated_0[(generated_0['visi'] >= 0) & (generated_0['visi'] < 100)]
+    well_generated1 = generated_1[(generated_1['visi'] >= 100) & (generated_1['visi'] < 500)]
+    smote_gan_data = pd.concat([lerp_data, well_generated0, well_generated1], axis=0)
+    # 제거변수 복구
+    smote_gan_data['binary_class'] = smote_gan_data['multi_class'].apply(lambda x: 0 if x == 2 else 1)
+    smote_gan_data['hour_sin'] = np.sin(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['hour_cos'] = np.cos(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['month_sin'] = np.sin(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['month_cos'] = np.cos(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['ground_temp - temp_C'] = smote_gan_data['groundtemp'] - smote_gan_data['temp_C']
+
+    filtered_data = smote_gan_data[smote_gan_data['multi_class'] != 2]
+    original_class2 = data[data['multi_class'] == 2]
+    final_data = pd.concat([filtered_data, original_class2], axis=0)
+    final_data.reset_index(drop=True, inplace=True)
+
+    # 결과 저장
+    final_data.to_csv(output_path, index = False)
+    print(len(final_data[final_data['multi_class']==0]),'|',len(final_data[final_data['multi_class']==1]),'|',len(final_data[final_data['multi_class']==2]))

Analysis_code/make_oversample_data/gan_sample_20000_2.py

@@ -0,0 +1,183 @@
+import pandas as pd
+import numpy as np
+import os
+os.environ["CUDA_VISIBLE_DEVICES"] = "1"
+from imblearn.over_sampling import SMOTENC
+import optuna
+from ctgan import CTGAN
+import torch
+import warnings
+
+# 지역별 데이터 파일 경로
+# regions = ['busan', 'daegu', 'daejeon', 'gwangju', 'incheon', 'seoul']
+regions = ['incheon', 'seoul','busan', 'daegu', 'daejeon', 'gwangju']
+file_paths = [f'../../data/data_for_modeling/{region}_train.csv' for region in regions]
+output_paths = [f'../../data/data_oversampled/ctgan20000/ctgan20000_2_{region}.csv' for region in regions]
+
+# GPU 사용 설정
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+print(f"Using device: {device}")
+
+# 경고 무시
+warnings.filterwarnings("ignore", category=UserWarning, module="optuna.distributions")
+
+# 지역별 처리
+for file_path, output_path in zip(file_paths, output_paths):
+    # 데이터 로드
+    data = pd.read_csv(file_path, index_col=0)
+    data= data.loc[data['year'].isin([2018,2020]),:]
+    data['cloudcover'] = data['cloudcover'].astype('int')
+    data['lm_cloudcover'] = data['lm_cloudcover'].astype('int')
+    X = data.drop(columns=['multi_class', 'binary_class'])
+    y = data['multi_class']
+
+    # 불필요한 열 제거
+    X.drop(columns=['ground_temp - temp_C', 'hour_sin', 'hour_cos', 'month_sin', 'month_cos'], inplace=True)
+
+    # SMOTENC에서 사용할 범주형 변수 열 번호 설정
+    categorical_features_indices = [i for i, dtype in enumerate(X.dtypes) if dtype != 'float64']
+
+    # sampling_strategy 설정
+    count_class_0 = (y == 0).sum()
+    count_class_1 = (y == 1).sum()
+    count_class_2 = (y == 2).sum()
+    sampling_strategy = {
+        0: 500 if count_class_0 <= 500 else 1000,
+        1: int(np.ceil(count_class_1 / 100) * 100),  # 백의 자리로 올림
+        2: count_class_2
+    }
+
+    # SMOTENC 적용
+    smotenc = SMOTENC(categorical_features=categorical_features_indices, sampling_strategy=sampling_strategy, random_state=42)
+    X_resampled, y_resampled = smotenc.fit_resample(X, y)
+
+    # Resampled 데이터 생성
+    lerp_data = X_resampled.copy()
+    lerp_data['multi_class'] = y_resampled
+
+    # CTGAN에서 사용할 범주형 변수 열 이름 설정
+    categorical_features = [
+        col for col, dtype in zip(lerp_data.columns, lerp_data.dtypes) if dtype != 'float64'
+    ]
+
+    # Optuna 목적 함수 정의
+    def objective(trial):
+        # 하이퍼파라미터 탐색 범위 설정
+        embedding_dim = trial.suggest_int("embedding_dim", 64, 128)
+        generator_dim = trial.suggest_categorical("generator_dim", [(64, 64), (128, 128)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(64, 64), (128, 128)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [64, 128, 256])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 3)
+
+        # CTGAN 모델 생성
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        # 범주 0 데이터 필터링
+        data_0 = lerp_data[lerp_data['multi_class'] == 0]
+
+        # 모델 학습
+        ctgan.fit(data_0, discrete_columns=categorical_features)
+
+        # 샘플 생성
+        generated_data = ctgan.sample(len(data_0) * 2)
+
+        # 평가: 샘플의 연속형 변수 분포 비교
+        real_visi = data_0['visi']
+        generated_visi = generated_data['visi']
+
+        # 분포 간 차이(MSE) 계산
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    # Optuna로 최적화 수행
+    study = optuna.create_study(direction="maximize")
+    study.optimize(objective, n_trials=50)
+
+    # 최적 하이퍼파라미터 출력
+    best_params = study.best_params
+
+    # 최적 하이퍼파라미터로 CTGAN 학습 및 샘플 생성
+    ctgan = CTGAN(
+        embedding_dim=best_params["embedding_dim"],
+        generator_dim=best_params["generator_dim"],
+        discriminator_dim=best_params["discriminator_dim"],
+        batch_size=best_params["batch_size"],
+        discriminator_steps=best_params["discriminator_steps"],
+        pac=best_params["pac"]
+    )
+
+    # 범주 0 데이터로 최종 학습
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 0], discrete_columns=categorical_features)
+    generated_0 = ctgan.sample(20000)
+
+    # 범주 1 데이터 최적화 및 생성
+    def objective_class1(trial):
+        embedding_dim = trial.suggest_int("embedding_dim", 128, 512)
+        generator_dim = trial.suggest_categorical("generator_dim", [(128, 128), (256, 256)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(128, 128), (256, 256)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [256, 512, 1024])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 5)
+
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        data_1 = lerp_data[lerp_data['multi_class'] == 1]
+        ctgan.fit(data_1, discrete_columns=categorical_features)
+        generated_data = ctgan.sample(len(data_1) * 2)
+
+        real_visi = data_1['visi']
+        generated_visi = generated_data['visi']
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    study_class1 = optuna.create_study(direction="maximize")
+    study_class1.optimize(objective_class1, n_trials=30)
+
+    best_params_class1 = study_class1.best_params
+    ctgan = CTGAN(
+        embedding_dim=best_params_class1["embedding_dim"],
+        generator_dim=best_params_class1["generator_dim"],
+        discriminator_dim=best_params_class1["discriminator_dim"],
+        batch_size=best_params_class1["batch_size"],
+        discriminator_steps=best_params_class1["discriminator_steps"],
+        pac=best_params_class1["pac"]
+    )
+
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 1], discrete_columns=categorical_features)
+    generated_1 = ctgan.sample(20000 - int(np.ceil(count_class_1 / 100) * 100))
+
+    # 데이터 병합 및 저장
+    well_generated0 = generated_0[(generated_0['visi'] >= 0) & (generated_0['visi'] < 100)]
+    well_generated1 = generated_1[(generated_1['visi'] >= 100) & (generated_1['visi'] < 500)]
+    smote_gan_data = pd.concat([lerp_data, well_generated0, well_generated1], axis=0)
+    # 제거변수 복구
+    smote_gan_data['binary_class'] = smote_gan_data['multi_class'].apply(lambda x: 0 if x == 2 else 1)
+    smote_gan_data['hour_sin'] = np.sin(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['hour_cos'] = np.cos(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['month_sin'] = np.sin(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['month_cos'] = np.cos(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['ground_temp - temp_C'] = smote_gan_data['groundtemp'] - smote_gan_data['temp_C']
+
+    filtered_data = smote_gan_data[smote_gan_data['multi_class'] != 2]
+    original_class2 = data[data['multi_class'] == 2]
+    final_data = pd.concat([filtered_data, original_class2], axis=0)
+    final_data.reset_index(drop=True, inplace=True)
+
+    # 결과 저장
+    final_data.to_csv(output_path, index = False)
+    print(len(final_data[final_data['multi_class']==0]),'|',len(final_data[final_data['multi_class']==1]),'|',len(final_data[final_data['multi_class']==2]))

Analysis_code/make_oversample_data/gan_sample_20000_3.py

@@ -0,0 +1,183 @@
+import pandas as pd
+import numpy as np
+import os
+os.environ["CUDA_VISIBLE_DEVICES"] = "1"
+from imblearn.over_sampling import SMOTENC
+import optuna
+from ctgan import CTGAN
+import torch
+import warnings
+
+# 지역별 데이터 파일 경로
+# regions = ['busan', 'daegu', 'daejeon', 'gwangju', 'incheon', 'seoul']
+regions = ['incheon', 'seoul','busan', 'daegu', 'daejeon', 'gwangju']
+file_paths = [f'../../data/data_for_modeling/{region}_train.csv' for region in regions]
+output_paths = [f'../../data/data_oversampled/ctgan20000/ctgan20000_3_{region}.csv' for region in regions]
+
+# GPU 사용 설정
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+print(f"Using device: {device}")
+
+# 경고 무시
+warnings.filterwarnings("ignore", category=UserWarning, module="optuna.distributions")
+
+# 지역별 처리
+for file_path, output_path in zip(file_paths, output_paths):
+    # 데이터 로드
+    data = pd.read_csv(file_path, index_col=0)
+    data= data.loc[data['year'].isin([2019,2020]),:]
+    data['cloudcover'] = data['cloudcover'].astype('int')
+    data['lm_cloudcover'] = data['lm_cloudcover'].astype('int')
+    X = data.drop(columns=['multi_class', 'binary_class'])
+    y = data['multi_class']
+
+    # 불필요한 열 제거
+    X.drop(columns=['ground_temp - temp_C', 'hour_sin', 'hour_cos', 'month_sin', 'month_cos'], inplace=True)
+
+    # SMOTENC에서 사용할 범주형 변수 열 번호 설정
+    categorical_features_indices = [i for i, dtype in enumerate(X.dtypes) if dtype != 'float64']
+
+    # sampling_strategy 설정
+    count_class_0 = (y == 0).sum()
+    count_class_1 = (y == 1).sum()
+    count_class_2 = (y == 2).sum()
+    sampling_strategy = {
+        0: 500 if count_class_0 <= 500 else 1000,
+        1: int(np.ceil(count_class_1 / 100) * 100),  # 백의 자리로 올림
+        2: count_class_2
+    }
+
+    # SMOTENC 적용
+    smotenc = SMOTENC(categorical_features=categorical_features_indices, sampling_strategy=sampling_strategy, random_state=42)
+    X_resampled, y_resampled = smotenc.fit_resample(X, y)
+
+    # Resampled 데이터 생성
+    lerp_data = X_resampled.copy()
+    lerp_data['multi_class'] = y_resampled
+
+    # CTGAN에서 사용할 범주형 변수 열 이름 설정
+    categorical_features = [
+        col for col, dtype in zip(lerp_data.columns, lerp_data.dtypes) if dtype != 'float64'
+    ]
+
+    # Optuna 목적 함수 정의
+    def objective(trial):
+        # 하이퍼파라미터 탐색 범위 설정
+        embedding_dim = trial.suggest_int("embedding_dim", 64, 128)
+        generator_dim = trial.suggest_categorical("generator_dim", [(64, 64), (128, 128)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(64, 64), (128, 128)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [64, 128, 256])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 3)
+
+        # CTGAN 모델 생성
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        # 범주 0 데이터 필터링
+        data_0 = lerp_data[lerp_data['multi_class'] == 0]
+
+        # 모델 학습
+        ctgan.fit(data_0, discrete_columns=categorical_features)
+
+        # 샘플 생성
+        generated_data = ctgan.sample(len(data_0) * 2)
+
+        # 평가: 샘플의 연속형 변수 분포 비교
+        real_visi = data_0['visi']
+        generated_visi = generated_data['visi']
+
+        # 분포 간 차이(MSE) 계산
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    # Optuna로 최적화 수행
+    study = optuna.create_study(direction="maximize")
+    study.optimize(objective, n_trials=50)
+
+    # 최적 하이퍼파라미터 출력
+    best_params = study.best_params
+
+    # 최적 하이퍼파라미터로 CTGAN 학습 및 샘플 생성
+    ctgan = CTGAN(
+        embedding_dim=best_params["embedding_dim"],
+        generator_dim=best_params["generator_dim"],
+        discriminator_dim=best_params["discriminator_dim"],
+        batch_size=best_params["batch_size"],
+        discriminator_steps=best_params["discriminator_steps"],
+        pac=best_params["pac"]
+    )
+
+    # 범주 0 데이터로 최종 학습
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 0], discrete_columns=categorical_features)
+    generated_0 = ctgan.sample(20000)
+
+    # 범주 1 데이터 최적화 및 생성
+    def objective_class1(trial):
+        embedding_dim = trial.suggest_int("embedding_dim", 128, 512)
+        generator_dim = trial.suggest_categorical("generator_dim", [(128, 128), (256, 256)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(128, 128), (256, 256)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [256, 512, 1024])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 5)
+
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        data_1 = lerp_data[lerp_data['multi_class'] == 1]
+        ctgan.fit(data_1, discrete_columns=categorical_features)
+        generated_data = ctgan.sample(len(data_1) * 2)
+
+        real_visi = data_1['visi']
+        generated_visi = generated_data['visi']
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    study_class1 = optuna.create_study(direction="maximize")
+    study_class1.optimize(objective_class1, n_trials=30)
+
+    best_params_class1 = study_class1.best_params
+    ctgan = CTGAN(
+        embedding_dim=best_params_class1["embedding_dim"],
+        generator_dim=best_params_class1["generator_dim"],
+        discriminator_dim=best_params_class1["discriminator_dim"],
+        batch_size=best_params_class1["batch_size"],
+        discriminator_steps=best_params_class1["discriminator_steps"],
+        pac=best_params_class1["pac"]
+    )
+
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 1], discrete_columns=categorical_features)
+    generated_1 = ctgan.sample(20000 - int(np.ceil(count_class_1 / 100) * 100))
+
+    # 데이터 병합 및 저장
+    well_generated0 = generated_0[(generated_0['visi'] >= 0) & (generated_0['visi'] < 100)]
+    well_generated1 = generated_1[(generated_1['visi'] >= 100) & (generated_1['visi'] < 500)]
+    smote_gan_data = pd.concat([lerp_data, well_generated0, well_generated1], axis=0)
+    # 제거변수 복구
+    smote_gan_data['binary_class'] = smote_gan_data['multi_class'].apply(lambda x: 0 if x == 2 else 1)
+    smote_gan_data['hour_sin'] = np.sin(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['hour_cos'] = np.cos(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['month_sin'] = np.sin(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['month_cos'] = np.cos(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['ground_temp - temp_C'] = smote_gan_data['groundtemp'] - smote_gan_data['temp_C']
+
+    filtered_data = smote_gan_data[smote_gan_data['multi_class'] != 2]
+    original_class2 = data[data['multi_class'] == 2]
+    final_data = pd.concat([filtered_data, original_class2], axis=0)
+    final_data.reset_index(drop=True, inplace=True)
+
+    # 결과 저장
+    final_data.to_csv(output_path, index = False)
+    print(len(final_data[final_data['multi_class']==0]),'|',len(final_data[final_data['multi_class']==1]),'|',len(final_data[final_data['multi_class']==2]))

Analysis_code/make_oversample_data/gan_sample_7000_1.py

@@ -0,0 +1,180 @@
+import pandas as pd
+import numpy as np
+from imblearn.over_sampling import SMOTENC
+import optuna
+from ctgan import CTGAN
+import torch
+import warnings
+
+# 지역별 데이터 파일 경로
+regions = ['incheon', 'seoul','busan', 'daegu', 'daejeon', 'gwangju']
+file_paths = [f'../../data/data_for_modeling/{region}_train.csv' for region in regions]
+output_paths = [f'../../data/data_oversampled/ctgan7000/ctgan7000_1_{region}.csv' for region in regions]
+
+# GPU 사용 설정
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+print(f"Using device: {device}")
+
+# 경고 무시
+warnings.filterwarnings("ignore", category=UserWarning, module="optuna.distributions")
+
+# 지역별 처리
+for file_path, output_path in zip(file_paths, output_paths):
+    # 데이터 로드
+    data = pd.read_csv(file_path, index_col=0)
+    data= data.loc[data['year'].isin([2018,2019]),:]
+    data['cloudcover'] = data['cloudcover'].astype('int')
+    data['lm_cloudcover'] = data['lm_cloudcover'].astype('int')
+    X = data.drop(columns=['multi_class', 'binary_class'])
+    y = data['multi_class']
+
+    # 불필요한 열 제거
+    X.drop(columns=['ground_temp - temp_C', 'hour_sin', 'hour_cos', 'month_sin', 'month_cos'], inplace=True)
+
+    # SMOTENC에서 사용할 범주형 변수 열 번호 설정
+    categorical_features_indices = [i for i, dtype in enumerate(X.dtypes) if dtype != 'float64']
+
+    # sampling_strategy 설정
+    count_class_0 = (y == 0).sum()
+    count_class_1 = (y == 1).sum()
+    count_class_2 = (y == 2).sum()
+    sampling_strategy = {
+        0: 500 if count_class_0 <= 500 else 1000,
+        1: int(np.ceil(count_class_1 / 100) * 100),  # 백의 자리로 올림
+        2: count_class_2
+    }
+
+    # SMOTENC 적용
+    smotenc = SMOTENC(categorical_features=categorical_features_indices, sampling_strategy=sampling_strategy, random_state=42)
+    X_resampled, y_resampled = smotenc.fit_resample(X, y)
+
+    # Resampled 데이터 생성
+    lerp_data = X_resampled.copy()
+    lerp_data['multi_class'] = y_resampled
+
+    # CTGAN에서 사용할 범주형 변수 열 이름 설정
+    categorical_features = [
+        col for col, dtype in zip(lerp_data.columns, lerp_data.dtypes) if dtype != 'float64'
+    ]
+
+    # Optuna 목적 함수 정의
+    def objective(trial):
+        # 하이퍼파라미터 탐색 범위 설정
+        embedding_dim = trial.suggest_int("embedding_dim", 64, 128)
+        generator_dim = trial.suggest_categorical("generator_dim", [(64, 64), (128, 128)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(64, 64), (128, 128)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [64, 128, 256])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 3)
+
+        # CTGAN 모델 생성
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        # 범주 0 데이터 필터링
+        data_0 = lerp_data[lerp_data['multi_class'] == 0]
+
+        # 모델 학습
+        ctgan.fit(data_0, discrete_columns=categorical_features)
+
+        # 샘플 생성
+        generated_data = ctgan.sample(len(data_0) * 2)
+
+        # 평가: 샘플의 연속형 변수 분포 비교
+        real_visi = data_0['visi']
+        generated_visi = generated_data['visi']
+
+        # 분포 간 차이(MSE) 계산
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    # Optuna로 최적화 수행
+    study = optuna.create_study(direction="maximize")
+    study.optimize(objective, n_trials=50)
+
+    # 최적 하이퍼파라미터 출력
+    best_params = study.best_params
+
+    # 최적 하이퍼파라미터로 CTGAN 학습 및 샘플 생성
+    ctgan = CTGAN(
+        embedding_dim=best_params["embedding_dim"],
+        generator_dim=best_params["generator_dim"],
+        discriminator_dim=best_params["discriminator_dim"],
+        batch_size=best_params["batch_size"],
+        discriminator_steps=best_params["discriminator_steps"],
+        pac=best_params["pac"]
+    )
+
+    # 범주 0 데이터로 최종 학습
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 0], discrete_columns=categorical_features)
+    generated_0 = ctgan.sample(7000)
+
+    # 범주 1 데이터 최적화 및 생성
+    def objective_class1(trial):
+        embedding_dim = trial.suggest_int("embedding_dim", 128, 512)
+        generator_dim = trial.suggest_categorical("generator_dim", [(128, 128), (256, 256)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(128, 128), (256, 256)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [256, 512, 1024])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 5)
+
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        data_1 = lerp_data[lerp_data['multi_class'] == 1]
+        ctgan.fit(data_1, discrete_columns=categorical_features)
+        generated_data = ctgan.sample(len(data_1) * 2)
+
+        real_visi = data_1['visi']
+        generated_visi = generated_data['visi']
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    study_class1 = optuna.create_study(direction="maximize")
+    study_class1.optimize(objective_class1, n_trials=30)
+
+    best_params_class1 = study_class1.best_params
+    ctgan = CTGAN(
+        embedding_dim=best_params_class1["embedding_dim"],
+        generator_dim=best_params_class1["generator_dim"],
+        discriminator_dim=best_params_class1["discriminator_dim"],
+        batch_size=best_params_class1["batch_size"],
+        discriminator_steps=best_params_class1["discriminator_steps"],
+        pac=best_params_class1["pac"]
+    )
+
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 1], discrete_columns=categorical_features)
+    generated_1 = ctgan.sample(7000 - int(np.ceil(count_class_1 / 100) * 100))
+
+    # 데이터 병합 및 저장
+    well_generated0 = generated_0[(generated_0['visi'] >= 0) & (generated_0['visi'] < 100)]
+    well_generated1 = generated_1[(generated_1['visi'] >= 100) & (generated_1['visi'] < 500)]
+    smote_gan_data = pd.concat([lerp_data, well_generated0, well_generated1], axis=0)
+    # 제거변수 복구
+    smote_gan_data['binary_class'] = smote_gan_data['multi_class'].apply(lambda x: 0 if x == 2 else 1)
+    smote_gan_data['hour_sin'] = np.sin(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['hour_cos'] = np.cos(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['month_sin'] = np.sin(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['month_cos'] = np.cos(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['ground_temp - temp_C'] = smote_gan_data['groundtemp'] - smote_gan_data['temp_C']
+
+    filtered_data = smote_gan_data[smote_gan_data['multi_class'] != 2]
+    original_class2 = data[data['multi_class'] == 2]
+    final_data = pd.concat([filtered_data, original_class2], axis=0)
+    final_data.reset_index(drop=True, inplace=True)
+
+    # 결과 저장
+    final_data.to_csv(output_path, index = False)
+    print(len(final_data[final_data['multi_class']==0]),'|',len(final_data[final_data['multi_class']==1]),'|',len(final_data[final_data['multi_class']==2]))

Analysis_code/make_oversample_data/gan_sample_7000_2.py

@@ -0,0 +1,182 @@
+import pandas as pd
+import numpy as np
+import os
+os.environ["CUDA_VISIBLE_DEVICES"] = "1"
+from imblearn.over_sampling import SMOTENC
+import optuna
+from ctgan import CTGAN
+import torch
+import warnings
+
+# 지역별 데이터 파일 경로
+regions = ['incheon', 'seoul','busan', 'daegu', 'daejeon', 'gwangju']
+file_paths = [f'../../data/data_for_modeling/{region}_train.csv' for region in regions]
+output_paths = [f'../../data/data_oversampled/ctgan7000/ctgan7000_2_{region}.csv' for region in regions]
+
+# GPU 사용 설정
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+print(f"Using device: {device}")
+
+# 경고 무시
+warnings.filterwarnings("ignore", category=UserWarning, module="optuna.distributions")
+
+# 지역별 처리
+for file_path, output_path in zip(file_paths, output_paths):
+    # 데이터 로드
+    data = pd.read_csv(file_path, index_col=0)
+    data= data.loc[data['year'].isin([2018,2020]),:]
+    data['cloudcover'] = data['cloudcover'].astype('int')
+    data['lm_cloudcover'] = data['lm_cloudcover'].astype('int')
+    X = data.drop(columns=['multi_class', 'binary_class'])
+    y = data['multi_class']
+
+    # 불필요한 열 제거
+    X.drop(columns=['ground_temp - temp_C', 'hour_sin', 'hour_cos', 'month_sin', 'month_cos'], inplace=True)
+
+    # SMOTENC에서 사용할 범주형 변수 열 번호 설정
+    categorical_features_indices = [i for i, dtype in enumerate(X.dtypes) if dtype != 'float64']
+
+    # sampling_strategy 설정
+    count_class_0 = (y == 0).sum()
+    count_class_1 = (y == 1).sum()
+    count_class_2 = (y == 2).sum()
+    sampling_strategy = {
+        0: 500 if count_class_0 <= 500 else 1000,
+        1: int(np.ceil(count_class_1 / 100) * 100),  # 백의 자리로 올림
+        2: count_class_2
+    }
+
+    # SMOTENC 적용
+    smotenc = SMOTENC(categorical_features=categorical_features_indices, sampling_strategy=sampling_strategy, random_state=42)
+    X_resampled, y_resampled = smotenc.fit_resample(X, y)
+
+    # Resampled 데이터 생성
+    lerp_data = X_resampled.copy()
+    lerp_data['multi_class'] = y_resampled
+
+    # CTGAN에서 사용할 범주형 변수 열 이름 설정
+    categorical_features = [
+        col for col, dtype in zip(lerp_data.columns, lerp_data.dtypes) if dtype != 'float64'
+    ]
+
+    # Optuna 목적 함수 정의
+    def objective(trial):
+        # 하이퍼파라미터 탐색 범위 설정
+        embedding_dim = trial.suggest_int("embedding_dim", 64, 128)
+        generator_dim = trial.suggest_categorical("generator_dim", [(64, 64), (128, 128)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(64, 64), (128, 128)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [64, 128, 256])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 3)
+
+        # CTGAN 모델 생성
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        # 범주 0 데이터 필터링
+        data_0 = lerp_data[lerp_data['multi_class'] == 0]
+
+        # 모델 학습
+        ctgan.fit(data_0, discrete_columns=categorical_features)
+
+        # 샘플 생성
+        generated_data = ctgan.sample(len(data_0) * 2)
+
+        # 평가: 샘플의 연속형 변수 분포 비교
+        real_visi = data_0['visi']
+        generated_visi = generated_data['visi']
+
+        # 분포 간 차이(MSE) 계산
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    # Optuna로 최적화 수행
+    study = optuna.create_study(direction="maximize")
+    study.optimize(objective, n_trials=50)
+
+    # 최적 하이퍼파라미터 출력
+    best_params = study.best_params
+
+    # 최적 하이퍼파라미터로 CTGAN 학습 및 샘플 생성
+    ctgan = CTGAN(
+        embedding_dim=best_params["embedding_dim"],
+        generator_dim=best_params["generator_dim"],
+        discriminator_dim=best_params["discriminator_dim"],
+        batch_size=best_params["batch_size"],
+        discriminator_steps=best_params["discriminator_steps"],
+        pac=best_params["pac"]
+    )
+
+    # 범주 0 데이터로 최종 학습
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 0], discrete_columns=categorical_features)
+    generated_0 = ctgan.sample(7000)
+
+    # 범주 1 데이터 최적화 및 생성
+    def objective_class1(trial):
+        embedding_dim = trial.suggest_int("embedding_dim", 128, 512)
+        generator_dim = trial.suggest_categorical("generator_dim", [(128, 128), (256, 256)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(128, 128), (256, 256)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [256, 512, 1024])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 5)
+
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        data_1 = lerp_data[lerp_data['multi_class'] == 1]
+        ctgan.fit(data_1, discrete_columns=categorical_features)
+        generated_data = ctgan.sample(len(data_1) * 2)
+
+        real_visi = data_1['visi']
+        generated_visi = generated_data['visi']
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    study_class1 = optuna.create_study(direction="maximize")
+    study_class1.optimize(objective_class1, n_trials=30)
+
+    best_params_class1 = study_class1.best_params
+    ctgan = CTGAN(
+        embedding_dim=best_params_class1["embedding_dim"],
+        generator_dim=best_params_class1["generator_dim"],
+        discriminator_dim=best_params_class1["discriminator_dim"],
+        batch_size=best_params_class1["batch_size"],
+        discriminator_steps=best_params_class1["discriminator_steps"],
+        pac=best_params_class1["pac"]
+    )
+
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 1], discrete_columns=categorical_features)
+    generated_1 = ctgan.sample(7000 - int(np.ceil(count_class_1 / 100) * 100))
+
+    # 데이터 병합 및 저장
+    well_generated0 = generated_0[(generated_0['visi'] >= 0) & (generated_0['visi'] < 100)]
+    well_generated1 = generated_1[(generated_1['visi'] >= 100) & (generated_1['visi'] < 500)]
+    smote_gan_data = pd.concat([lerp_data, well_generated0, well_generated1], axis=0)
+    # 제거변수 복구
+    smote_gan_data['binary_class'] = smote_gan_data['multi_class'].apply(lambda x: 0 if x == 2 else 1)
+    smote_gan_data['hour_sin'] = np.sin(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['hour_cos'] = np.cos(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['month_sin'] = np.sin(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['month_cos'] = np.cos(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['ground_temp - temp_C'] = smote_gan_data['groundtemp'] - smote_gan_data['temp_C']
+
+    filtered_data = smote_gan_data[smote_gan_data['multi_class'] != 2]
+    original_class2 = data[data['multi_class'] == 2]
+    final_data = pd.concat([filtered_data, original_class2], axis=0)
+    final_data.reset_index(drop=True, inplace=True)
+
+    # 결과 저장
+    final_data.to_csv(output_path, index = False)
+    print(len(final_data[final_data['multi_class']==0]),'|',len(final_data[final_data['multi_class']==1]),'|',len(final_data[final_data['multi_class']==2]))

Analysis_code/make_oversample_data/gan_sample_7000_3.py

@@ -0,0 +1,182 @@
+import pandas as pd
+import numpy as np
+import os
+os.environ["CUDA_VISIBLE_DEVICES"] = "0"
+from imblearn.over_sampling import SMOTENC
+import optuna
+from ctgan import CTGAN
+import torch
+import warnings
+
+# 지역별 데이터 파일 경로
+regions = ['incheon', 'seoul','busan', 'daegu', 'daejeon', 'gwangju']
+file_paths = [f'../../data/data_for_modeling/{region}_train.csv' for region in regions]
+output_paths = [f'../../data/data_oversampled/ctgan7000/ctgan7000_3_{region}.csv' for region in regions]
+
+# GPU 사용 설정
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+print(f"Using device: {device}")
+
+# 경고 무시
+warnings.filterwarnings("ignore", category=UserWarning, module="optuna.distributions")
+
+# 지역별 처리
+for file_path, output_path in zip(file_paths, output_paths):
+    # 데이터 로드
+    data = pd.read_csv(file_path, index_col=0)
+    data= data.loc[data['year'].isin([2019,2020]),:]
+    data['cloudcover'] = data['cloudcover'].astype('int')
+    data['lm_cloudcover'] = data['lm_cloudcover'].astype('int')
+    X = data.drop(columns=['multi_class', 'binary_class'])
+    y = data['multi_class']
+
+    # 불필요한 열 제거
+    X.drop(columns=['ground_temp - temp_C', 'hour_sin', 'hour_cos', 'month_sin', 'month_cos'], inplace=True)
+
+    # SMOTENC에서 사용할 범주형 변수 열 번호 설정
+    categorical_features_indices = [i for i, dtype in enumerate(X.dtypes) if dtype != 'float64']
+
+    # sampling_strategy 설정
+    count_class_0 = (y == 0).sum()
+    count_class_1 = (y == 1).sum()
+    count_class_2 = (y == 2).sum()
+    sampling_strategy = {
+        0: 500 if count_class_0 <= 500 else 1000,
+        1: int(np.ceil(count_class_1 / 100) * 100),  # 백의 자리로 올림
+        2: count_class_2
+    }
+
+    # SMOTENC 적용
+    smotenc = SMOTENC(categorical_features=categorical_features_indices, sampling_strategy=sampling_strategy, random_state=42)
+    X_resampled, y_resampled = smotenc.fit_resample(X, y)
+
+    # Resampled 데이터 생성
+    lerp_data = X_resampled.copy()
+    lerp_data['multi_class'] = y_resampled
+
+    # CTGAN에서 사용할 범주형 변수 열 이름 설정
+    categorical_features = [
+        col for col, dtype in zip(lerp_data.columns, lerp_data.dtypes) if dtype != 'float64'
+    ]
+
+    # Optuna 목적 함수 정의
+    def objective(trial):
+        # 하이퍼파라미터 탐색 범위 설정
+        embedding_dim = trial.suggest_int("embedding_dim", 64, 128)
+        generator_dim = trial.suggest_categorical("generator_dim", [(64, 64), (128, 128)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(64, 64), (128, 128)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [64, 128, 256])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 3)
+
+        # CTGAN 모델 생성
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        # 범주 0 데이터 필터링
+        data_0 = lerp_data[lerp_data['multi_class'] == 0]
+
+        # 모델 학습
+        ctgan.fit(data_0, discrete_columns=categorical_features)
+
+        # 샘플 생성
+        generated_data = ctgan.sample(len(data_0) * 2)
+
+        # 평가: 샘플의 연속형 변수 분포 비교
+        real_visi = data_0['visi']
+        generated_visi = generated_data['visi']
+
+        # 분포 간 차이(MSE) 계산
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    # Optuna로 최적화 수행
+    study = optuna.create_study(direction="maximize")
+    study.optimize(objective, n_trials=50)
+
+    # 최적 하이퍼파라미터 출력
+    best_params = study.best_params
+
+    # 최적 하이퍼파라미터로 CTGAN 학습 및 샘플 생성
+    ctgan = CTGAN(
+        embedding_dim=best_params["embedding_dim"],
+        generator_dim=best_params["generator_dim"],
+        discriminator_dim=best_params["discriminator_dim"],
+        batch_size=best_params["batch_size"],
+        discriminator_steps=best_params["discriminator_steps"],
+        pac=best_params["pac"]
+    )
+
+    # 범주 0 데이터로 최종 학습
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 0], discrete_columns=categorical_features)
+    generated_0 = ctgan.sample(7000)
+
+    # 범주 1 데이터 최적화 및 생성
+    def objective_class1(trial):
+        embedding_dim = trial.suggest_int("embedding_dim", 128, 512)
+        generator_dim = trial.suggest_categorical("generator_dim", [(128, 128), (256, 256)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(128, 128), (256, 256)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [256, 512, 1024])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 5)
+
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        data_1 = lerp_data[lerp_data['multi_class'] == 1]
+        ctgan.fit(data_1, discrete_columns=categorical_features)
+        generated_data = ctgan.sample(len(data_1) * 2)
+
+        real_visi = data_1['visi']
+        generated_visi = generated_data['visi']
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    study_class1 = optuna.create_study(direction="maximize")
+    study_class1.optimize(objective_class1, n_trials=30)
+
+    best_params_class1 = study_class1.best_params
+    ctgan = CTGAN(
+        embedding_dim=best_params_class1["embedding_dim"],
+        generator_dim=best_params_class1["generator_dim"],
+        discriminator_dim=best_params_class1["discriminator_dim"],
+        batch_size=best_params_class1["batch_size"],
+        discriminator_steps=best_params_class1["discriminator_steps"],
+        pac=best_params_class1["pac"]
+    )
+
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 1], discrete_columns=categorical_features)
+    generated_1 = ctgan.sample(7000 - int(np.ceil(count_class_1 / 100) * 100))
+
+    # 데이터 병합 및 저장
+    well_generated0 = generated_0[(generated_0['visi'] >= 0) & (generated_0['visi'] < 100)]
+    well_generated1 = generated_1[(generated_1['visi'] >= 100) & (generated_1['visi'] < 500)]
+    smote_gan_data = pd.concat([lerp_data, well_generated0, well_generated1], axis=0)
+    # 제거변수 복구
+    smote_gan_data['binary_class'] = smote_gan_data['multi_class'].apply(lambda x: 0 if x == 2 else 1)
+    smote_gan_data['hour_sin'] = np.sin(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['hour_cos'] = np.cos(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['month_sin'] = np.sin(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['month_cos'] = np.cos(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['ground_temp - temp_C'] = smote_gan_data['groundtemp'] - smote_gan_data['temp_C']
+
+    filtered_data = smote_gan_data[smote_gan_data['multi_class'] != 2]
+    original_class2 = data[data['multi_class'] == 2]
+    final_data = pd.concat([filtered_data, original_class2], axis=0)
+    final_data.reset_index(drop=True, inplace=True)
+
+    # 결과 저장
+    final_data.to_csv(output_path, index = False)
+    print(len(final_data[final_data['multi_class']==0]),'|',len(final_data[final_data['multi_class']==1]),'|',len(final_data[final_data['multi_class']==2]))

Analysis_code/make_oversample_data/oversampling_code.py

@@ -0,0 +1,355 @@
+import pandas as pd
+import numpy as np
+from imblearn.over_sampling import SMOTENC
+
+
+# smote와 ctgan을 이용한 oversampling 진행
+
+# 파일 경로와 지역 이름 리스트
+regions = ['busan', 'daegu', 'daejeon', 'incheon', 'seoul','gwangju']
+input_paths = [f'../data/data_for_modeling/{region}_train.csv' for region in regions]
+
+
+
+
+# 반복적으로 각 지역 데이터 처리
+for region, input_path in zip(regions, input_paths):
+    # 데이터 읽기
+    data = pd.read_csv(input_path, index_col=0)
+    data.drop(['Unnamed: 0'], axis=1, inplace=True)
+    print("\n######",region,"#######")
+    print(len(data[data['multi_class']==0]),'|',len(data[data['multi_class']==1]),'|',len(data[data['multi_class']==2]))
+    print(len(data.columns))
+
+
+
+
+
+import pandas as pd
+import numpy as np
+from imblearn.over_sampling import SMOTENC
+
+# 파일 경로와 지역 이름 리스트
+regions = ['busan', 'daegu', 'daejeon', 'incheon', 'seoul','gwangju']
+input_paths = [f'../data/data_for_modeling/{region}_train.csv' for region in regions]
+output_paths = [f'../data/data_oversampled/smote_{region}.csv' for region in regions]
+
+# 반복적으로 각 지역 데이터 처리
+for region, input_path, output_path in zip(regions, input_paths, output_paths):
+    # 데이터 읽기
+    data = pd.read_csv(input_path, index_col=0)
+    data.drop(['Unnamed: 0'], axis=1, inplace=True)
+
+    # X와 y 분리
+    X = data.drop(columns=['multi_class', 'binary_class'])
+    y = data['multi_class']
+
+    # 불필요한 열 제거
+    X.drop(columns=['ground_temp - temp_C', 'hour_sin', 'hour_cos', 'month_sin', 'month_cos'], inplace=True)
+
+    # 범주형 변수 식별
+    categorical_features = [i for i, dtype in enumerate(X.dtypes) if dtype != 'float64']
+
+    # 각 지역의 multi_class 값이 2인 데이터 개수 계산
+    count_class_2 = (y == 2).sum()
+
+    # SMOTENC 적용
+    smotenc = SMOTENC(
+        categorical_features=categorical_features,
+        sampling_strategy={0: 10000, 1: 10000, 2: count_class_2},
+        random_state=42
+    )
+    X_resampled, y_resampled = smotenc.fit_resample(X, y)
+
+    # 추가 변수 생성
+    X_resampled['multi_class'] = y_resampled
+    X_resampled['binary_class'] = X_resampled['multi_class'].apply(lambda x: 0 if x == 2 else 1)
+    X_resampled['hour_sin'] = np.sin(2 * np.pi * X_resampled['hour'] / 24)
+    X_resampled['hour_cos'] = np.cos(2 * np.pi * X_resampled['hour'] / 24)
+    X_resampled['month_sin'] = np.sin(2 * np.pi * X_resampled['month'] / 12)
+    X_resampled['month_cos'] = np.cos(2 * np.pi * X_resampled['month'] / 12)
+    X_resampled['ground_temp - temp_C'] = X_resampled['groundtemp'] - X_resampled['temp_C']
+
+    # 결과 저장
+    X_resampled.to_csv(output_path)
+    print(f"Processed and saved: {region} -> {output_path}")
+
+
+smote_seoul = pd.read_csv('../data/data_oversampled/smote_seoul.csv')
+print(smote_seoul[smote_seoul['multi_class']==0]['visi'].describe())
+print(smote_seoul[smote_seoul['multi_class']==1]['visi'].describe())
+
+
+import pandas as pd
+import numpy as np
+from imblearn.over_sampling import SMOTENC
+
+# 파일 경로와 지역 이름 리스트
+regions = ['busan', 'daegu', 'daejeon', 'incheon', 'seoul','gwangju']
+input_paths = [f'../data/data_oversampled/smote_{region}.csv' for region in regions]
+
+# 반복적으로 각 지역 데이터 처리
+for region, input_path in zip(regions, input_paths):
+    # 데이터 읽기
+    data = pd.read_csv(input_path, index_col=0)
+    data.drop(['Unnamed: 0'], axis=1, inplace=True)
+    print("\n######",region,"#######")
+    print(len(data[data['multi_class']==0]),'|',len(data[data['multi_class']==1]),'|',len(data[data['multi_class']==2]))
+    print(len(data.columns))
+
+
+
+import pandas as pd
+import numpy as np
+from imblearn.over_sampling import SMOTENC
+
+# 파일 경로와 지역 이름 리스트
+regions = ['busan', 'daegu', 'daejeon', 'incheon', 'seoul','gwangju']
+input_paths = [f'../data/data_for_modeling/{region}_train.csv' for region in regions]
+
+# 반복적으로 각 지역 데이터 처리
+for region, input_path in zip(regions, input_paths):
+    # 데이터 읽기
+    data = pd.read_csv(input_path, index_col=0)
+    data.drop(['Unnamed: 0'], axis=1, inplace=True)
+    print("\n######",region,"#######")
+    print(len(data[data['multi_class']==0]),'|',len(data[data['multi_class']==1]),'|',len(data[data['multi_class']==2]))
+    print(len(data.columns))
+
+
+import pandas as pd
+import numpy as np
+from imblearn.over_sampling import SMOTENC
+import optuna
+from ctgan import CTGAN
+import torch
+import warnings
+
+# 지역별 데이터 파일 경로
+regions = ['busan', 'daegu', 'daejeon', 'incheon', 'seoul','gwangju']
+file_paths = [f'../data/data_for_modeling/df_{region}.feather' for region in regions]
+output_paths = [f'../data/data_oversampled/ctgan_{region}.csv' for region in regions]
+
+# GPU 사용 설정
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+print(f"Using device: {device}")
+
+# 경고 무시
+warnings.filterwarnings("ignore", category=UserWarning, module="optuna.distributions")
+
+# 지역별 처리
+for file_path, output_path in zip(file_paths, output_paths):
+    # 데이터 로드
+    data = pd.read_feather(file_path)
+    data.drop(['Unnamed: 0'], axis=1, inplace=True)
+    X = data.drop(columns=['multi_class', 'binary_class'])
+    y = data['multi_class']
+
+    # 불필요한 열 제거
+    X.drop(columns=['ground_temp - temp_C', 'hour_sin', 'hour_cos', 'month_sin', 'month_cos'], inplace=True)
+
+    # SMOTENC에서 사용할 범주형 변수 열 번호 설정
+    categorical_features_indices = [i for i, dtype in enumerate(X.dtypes) if dtype != 'float64']
+
+    # sampling_strategy 설정
+    count_class_0 = (y == 0).sum()
+    count_class_1 = (y == 1).sum()
+    count_class_2 = (y == 2).sum()
+    sampling_strategy = {
+        0: 500 if count_class_0 <= 500 else 1000,
+        1: int(np.ceil(count_class_1 / 100) * 100),  # 백의 자리로 올림
+        2: count_class_2
+    }
+
+    # SMOTENC 적용
+    smotenc = SMOTENC(categorical_features=categorical_features_indices, sampling_strategy=sampling_strategy, random_state=42)
+    X_resampled, y_resampled = smotenc.fit_resample(X, y)
+
+    # Resampled 데이터 생성
+    lerp_data = X_resampled.copy()
+    lerp_data['multi_class'] = y_resampled
+
+    # CTGAN에서 사용할 범주형 변수 열 이름 설정
+    categorical_features = [
+        col for col, dtype in zip(lerp_data.columns, lerp_data.dtypes) if dtype != 'float64'
+    ]
+
+    # Optuna 목적 함수 정의
+    def objective(trial):
+        # 하이퍼파라미터 탐색 범위 설정
+        embedding_dim = trial.suggest_int("embedding_dim", 64, 128)
+        generator_dim = trial.suggest_categorical("generator_dim", [(64, 64), (128, 128)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(64, 64), (128, 128)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [64, 128, 256])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 3)
+
+        # CTGAN 모델 생성
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        # 범주 0 데이터 필터링
+        data_0 = lerp_data[lerp_data['multi_class'] == 0]
+
+        # 모델 학습
+        ctgan.fit(data_0, discrete_columns=categorical_features)
+
+        # 샘플 생성
+        generated_data = ctgan.sample(len(data_0) * 2)
+
+        # 평가: 샘플의 연속형 변수 분포 비교
+        real_visi = data_0['visi']
+        generated_visi = generated_data['visi']
+
+        # 분포 간 차이(MSE) 계산
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    # Optuna로 최적화 수행
+    study = optuna.create_study(direction="maximize")
+    study.optimize(objective, n_trials=50)
+
+    # 최적 하이퍼파라미터 출력
+    best_params = study.best_params
+
+    # 최적 하이퍼파라미터로 CTGAN 학습 및 샘플 생성
+    ctgan = CTGAN(
+        embedding_dim=best_params["embedding_dim"],
+        generator_dim=best_params["generator_dim"],
+        discriminator_dim=best_params["discriminator_dim"],
+        batch_size=best_params["batch_size"],
+        discriminator_steps=best_params["discriminator_steps"],
+        pac=best_params["pac"]
+    )
+
+    # 범주 0 데이터로 최종 학습
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 0], discrete_columns=categorical_features)
+    generated_0 = ctgan.sample(19500 if count_class_0 <= 500 else 19000)
+
+    # 범주 1 데이터 최적화 및 생성
+    def objective_class1(trial):
+        embedding_dim = trial.suggest_int("embedding_dim", 128, 512)
+        generator_dim = trial.suggest_categorical("generator_dim", [(128, 128), (256, 256)])
+        discriminator_dim = trial.suggest_categorical("discriminator_dim", [(128, 128), (256, 256)])
+        pac = trial.suggest_categorical("pac", [4, 8])
+        batch_size = trial.suggest_categorical("batch_size", [256, 512, 1024])
+        discriminator_steps = trial.suggest_int("discriminator_steps", 1, 5)
+
+        ctgan = CTGAN(
+            embedding_dim=embedding_dim,
+            generator_dim=generator_dim,
+            discriminator_dim=discriminator_dim,
+            batch_size=batch_size,
+            discriminator_steps=discriminator_steps,
+            pac=pac
+        )
+
+        data_1 = lerp_data[lerp_data['multi_class'] == 1]
+        ctgan.fit(data_1, discrete_columns=categorical_features)
+        generated_data = ctgan.sample(len(data_1) * 2)
+
+        real_visi = data_1['visi']
+        generated_visi = generated_data['visi']
+        mse = ((real_visi.mean() - generated_visi.mean())**2 + (real_visi.std() - generated_visi.std())**2)
+        return -mse
+
+    study_class1 = optuna.create_study(direction="maximize")
+    study_class1.optimize(objective_class1, n_trials=30)
+
+    best_params_class1 = study_class1.best_params
+    ctgan = CTGAN(
+        embedding_dim=best_params_class1["embedding_dim"],
+        generator_dim=best_params_class1["generator_dim"],
+        discriminator_dim=best_params_class1["discriminator_dim"],
+        batch_size=best_params_class1["batch_size"],
+        discriminator_steps=best_params_class1["discriminator_steps"],
+        pac=best_params_class1["pac"]
+    )
+
+    ctgan.fit(lerp_data[lerp_data['multi_class'] == 1], discrete_columns=categorical_features)
+    generated_1 = ctgan.sample(20000 - int(np.ceil(count_class_1 / 100) * 100))
+
+    # 데이터 병합 및 저장
+    well_generated0 = generated_0[(generated_0['visi'] >= 0) & (generated_0['visi'] < 100)]
+    well_generated1 = generated_1[(generated_1['visi'] >= 100) & (generated_1['visi'] < 500)]
+    smote_gan_data = pd.concat([lerp_data, well_generated0, well_generated1], axis=0)
+        # 제거변수 복구
+    smote_gan_data['binary_class'] = smote_gan_data['multi_class'].apply(lambda x: 0 if x == 2 else 1)
+    smote_gan_data['hour_sin'] = np.sin(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['hour_cos'] = np.cos(2 * np.pi * smote_gan_data['hour'] / 24)
+    smote_gan_data['month_sin'] = np.sin(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['month_cos'] = np.cos(2 * np.pi * smote_gan_data['month'] / 12)
+    smote_gan_data['ground_temp - temp_C'] = smote_gan_data['groundtemp'] - smote_gan_data['temp_C']
+
+    # 결과 저장
+    smote_gan_data.to_csv(output_path, index = False)
+    print(f"Processed and saved: {region} -> {output_path}")
+
+
+
+import pandas as pd
+import numpy as np
+from imblearn.over_sampling import SMOTENC
+
+# 파일 경로와 지역 이름 리스트
+regions = ['busan', 'daegu', 'daejeon', 'incheon', 'seoul','gwangju']
+input_paths = [f'../data/data_oversampled/ctgan_{region}.csv' for region in regions]
+
+# 반복적으로 각 지역 데이터 처리
+for region, input_path in zip(regions, input_paths):
+    # 데이터 읽기
+    data = pd.read_csv(input_path)
+    print("\n######",region,"#######")
+    print(len(data[data['multi_class']==0]),'|',len(data[data['multi_class']==1]),'|',len(data[data['multi_class']==2]))
+    print(len(data.columns))
+
+
+
+
+busan_check = pd.read_csv('../data/data_oversampled/ctgan_busan.csv')
+print(busan_check[busan_check['multi_class']==0]['visi'].describe())
+print(busan_check[busan_check['multi_class']==1]['visi'].describe())
+print(busan_check[busan_check['multi_class']==2]['visi'].describe())
+
+
+
+
+import pandas as pd
+import numpy as np
+from imblearn.over_sampling import SMOTENC
+
+# 파일 경로와 지역 이름 리스트
+regions = ['busan', 'daegu', 'daejeon', 'incheon', 'seoul','gwangju']
+origin_paths = [f'../data/data_for_modeling/{region}_train.csv' for region in regions]
+augment_paths = [f'../data/data_oversampled/ctgan_{region}.csv' for region in regions]
+
+# 반복적으로 각 지역 데이터 처리
+for region, origin_path, augment_path in zip(regions, origin_paths, augment_paths):
+    # 데이터 읽기
+    origin = pd.read_csv(origin_path, index_col=0)
+    augment = pd.read_csv(augment_path)
+
+    # 증강된 데이터에서 범주 2 데이터 제거
+    filtered_data = augment[augment['multi_class'] != 2]
+
+    # 원본 데이터에서 범주 2 데이터 추출
+    original_class2 = origin[origin['multi_class'] == 2]
+
+    # 제거된 데이터에 원본 범주 2 데이터를 추가
+    final_data = pd.concat([filtered_data, original_class2], axis=0)
+
+    # 인덱스 재설정
+    final_data.reset_index(drop=True, inplace=True)
+
+    # 결과 저장
+    final_data.to_csv(augment_path, index = False)
+
+    print("\n######",region,"#######")
+    print(len(final_data[final_data['multi_class']==0]),'|',len(final_data[final_data['multi_class']==1]),'|',len(final_data[final_data['multi_class']==2]))
+    print(len(data.columns))

Analysis_code/make_oversample_data/smote_sample_1.py

@@ -0,0 +1,53 @@
+import pandas as pd
+import numpy as np
+
+from imblearn.over_sampling import SMOTENC
+
+# 지역별 데이터 파일 경로
+regions = ['incheon', 'seoul','busan', 'daegu', 'daejeon', 'gwangju']
+file_paths = [f'../../data/data_for_modeling/{region}_train.csv' for region in regions]
+output_paths = [f'../../data/data_oversampled/smote/smote_1_{region}.csv' for region in regions]
+
+# 지역별 처리
+for file_path, output_path in zip(file_paths, output_paths):
+    # 데이터 로드
+    data = pd.read_csv(file_path, index_col=0)
+    data= data.loc[data['year'].isin([2018,2019]),:]
+    data['cloudcover'] = data['cloudcover'].astype('int')
+    data['lm_cloudcover'] = data['lm_cloudcover'].astype('int')
+    X = data.drop(columns=['multi_class', 'binary_class'])
+    y = data['multi_class']
+
+    # 불필요한 열 제거
+    X.drop(columns=['ground_temp - temp_C', 'hour_sin', 'hour_cos', 'month_sin', 'month_cos'], inplace=True)
+
+    # SMOTENC에서 사용할 범주형 변수 열 번호 설정
+    categorical_features_indices = [i for i, dtype in enumerate(X.dtypes) if dtype != 'float64']
+
+    # sampling_strategy 설정
+    count_class_2 = (y == 2).sum()
+    sampling_strategy = {
+        0: int(np.ceil(count_class_2 / 1000) * 500),
+        1: int(np.ceil(count_class_2 / 1000) * 500),
+        2: count_class_2
+    }
+
+    # SMOTENC 적용
+    smotenc = SMOTENC(categorical_features=categorical_features_indices, sampling_strategy=sampling_strategy, random_state=42)
+    X_resampled, y_resampled = smotenc.fit_resample(X, y)
+
+    # Resampled 데이터 생성
+    lerp_data = X_resampled.copy()
+    lerp_data['multi_class'] = y_resampled
+
+    # 제거변수 복구
+    lerp_data['binary_class'] = lerp_data['multi_class'].apply(lambda x: 0 if x == 2 else 1)
+    lerp_data['hour_sin'] = np.sin(2 * np.pi * lerp_data['hour'] / 24)
+    lerp_data['hour_cos'] = np.cos(2 * np.pi * lerp_data['hour'] / 24)
+    lerp_data['month_sin'] = np.sin(2 * np.pi * lerp_data['month'] / 12)
+    lerp_data['month_cos'] = np.cos(2 * np.pi * lerp_data['month'] / 12)
+    lerp_data['ground_temp - temp_C'] = lerp_data['groundtemp'] - lerp_data['temp_C']
+
+    # 결과 저장
+    lerp_data.to_csv(output_path, index = False)
+    print(len(lerp_data[lerp_data['multi_class']==0]),'|',len(lerp_data[lerp_data['multi_class']==1]),'|',len(lerp_data[lerp_data['multi_class']==2]))

Analysis_code/make_oversample_data/smote_sample_2.py

@@ -0,0 +1,53 @@
+import pandas as pd
+import numpy as np
+
+from imblearn.over_sampling import SMOTENC
+
+# 지역별 데이터 파일 경로
+regions = ['incheon', 'seoul','busan', 'daegu', 'daejeon', 'gwangju']
+file_paths = [f'../../data/data_for_modeling/{region}_train.csv' for region in regions]
+output_paths = [f'../../data/data_oversampled/smote/smote_2_{region}.csv' for region in regions]
+
+# 지역별 처리
+for file_path, output_path in zip(file_paths, output_paths):
+    # 데이터 로드
+    data = pd.read_csv(file_path, index_col=0)
+    data= data.loc[data['year'].isin([2018,2020]),:]
+    data['cloudcover'] = data['cloudcover'].astype('int')
+    data['lm_cloudcover'] = data['lm_cloudcover'].astype('int')
+    X = data.drop(columns=['multi_class', 'binary_class'])
+    y = data['multi_class']
+
+    # 불필요한 열 제거
+    X.drop(columns=['ground_temp - temp_C', 'hour_sin', 'hour_cos', 'month_sin', 'month_cos'], inplace=True)
+
+    # SMOTENC에서 사용할 범주형 변수 열 번호 설정
+    categorical_features_indices = [i for i, dtype in enumerate(X.dtypes) if dtype != 'float64']
+
+    # sampling_strategy 설정
+    count_class_2 = (y == 2).sum()
+    sampling_strategy = {
+        0: int(np.ceil(count_class_2 / 1000) * 500),
+        1: int(np.ceil(count_class_2 / 1000) * 500),
+        2: count_class_2
+    }
+
+    # SMOTENC 적용
+    smotenc = SMOTENC(categorical_features=categorical_features_indices, sampling_strategy=sampling_strategy, random_state=42)
+    X_resampled, y_resampled = smotenc.fit_resample(X, y)
+
+    # Resampled 데이터 생성
+    lerp_data = X_resampled.copy()
+    lerp_data['multi_class'] = y_resampled
+
+    # 제거변수 복구
+    lerp_data['binary_class'] = lerp_data['multi_class'].apply(lambda x: 0 if x == 2 else 1)
+    lerp_data['hour_sin'] = np.sin(2 * np.pi * lerp_data['hour'] / 24)
+    lerp_data['hour_cos'] = np.cos(2 * np.pi * lerp_data['hour'] / 24)
+    lerp_data['month_sin'] = np.sin(2 * np.pi * lerp_data['month'] / 12)
+    lerp_data['month_cos'] = np.cos(2 * np.pi * lerp_data['month'] / 12)
+    lerp_data['ground_temp - temp_C'] = lerp_data['groundtemp'] - lerp_data['temp_C']
+
+    # 결과 저장
+    lerp_data.to_csv(output_path, index = False)
+    print(len(lerp_data[lerp_data['multi_class']==0]),'|',len(lerp_data[lerp_data['multi_class']==1]),'|',len(lerp_data[lerp_data['multi_class']==2]))

Analysis_code/make_oversample_data/smote_sample_3.py

@@ -0,0 +1,53 @@
+import pandas as pd
+import numpy as np
+
+from imblearn.over_sampling import SMOTENC
+
+# 지역별 데이터 파일 경로
+regions = ['incheon', 'seoul','busan', 'daegu', 'daejeon', 'gwangju']
+file_paths = [f'../../data/data_for_modeling/{region}_train.csv' for region in regions]
+output_paths = [f'../../data/data_oversampled/smote/smote_3_{region}.csv' for region in regions]
+
+# 지역별 처리
+for file_path, output_path in zip(file_paths, output_paths):
+    # 데이터 로드
+    data = pd.read_csv(file_path, index_col=0)
+    data= data.loc[data['year'].isin([2019,2020]),:]
+    data['cloudcover'] = data['cloudcover'].astype('int')
+    data['lm_cloudcover'] = data['lm_cloudcover'].astype('int')
+    X = data.drop(columns=['multi_class', 'binary_class'])
+    y = data['multi_class']
+
+    # 불필요한 열 제거
+    X.drop(columns=['ground_temp - temp_C', 'hour_sin', 'hour_cos', 'month_sin', 'month_cos'], inplace=True)
+
+    # SMOTENC에서 사용할 범주형 변수 열 번호 설정
+    categorical_features_indices = [i for i, dtype in enumerate(X.dtypes) if dtype != 'float64']
+
+    # sampling_strategy 설정
+    count_class_2 = (y == 2).sum()
+    sampling_strategy = {
+        0: int(np.ceil(count_class_2 / 1000) * 500),
+        1: int(np.ceil(count_class_2 / 1000) * 500),
+        2: count_class_2
+    }
+
+    # SMOTENC 적용
+    smotenc = SMOTENC(categorical_features=categorical_features_indices, sampling_strategy=sampling_strategy, random_state=42)
+    X_resampled, y_resampled = smotenc.fit_resample(X, y)
+
+    # Resampled 데이터 생성
+    lerp_data = X_resampled.copy()
+    lerp_data['multi_class'] = y_resampled
+
+    # 제거변수 복구
+    lerp_data['binary_class'] = lerp_data['multi_class'].apply(lambda x: 0 if x == 2 else 1)
+    lerp_data['hour_sin'] = np.sin(2 * np.pi * lerp_data['hour'] / 24)
+    lerp_data['hour_cos'] = np.cos(2 * np.pi * lerp_data['hour'] / 24)
+    lerp_data['month_sin'] = np.sin(2 * np.pi * lerp_data['month'] / 12)
+    lerp_data['month_cos'] = np.cos(2 * np.pi * lerp_data['month'] / 12)
+    lerp_data['ground_temp - temp_C'] = lerp_data['groundtemp'] - lerp_data['temp_C']
+
+    # 결과 저장
+    lerp_data.to_csv(output_path, index = False)
+    print(len(lerp_data[lerp_data['multi_class']==0]),'|',len(lerp_data[lerp_data['multi_class']==1]),'|',len(lerp_data[lerp_data['multi_class']==2]))

Analysis_code/make_train_test.ipynb

@@ -0,0 +1,1099 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": 1,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import pandas as pd\n",
+    "import numpy as np\n",
+    "import matplotlib.pyplot as plt\n",
+    "import seaborn as sns\n",
+    "from sklearn.model_selection import train_test_split\n",
+    "from collections import Counter"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 2,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df_seoul = pd.read_feather(\"../data/data_for_modeling/df_seoul.feather\")\n",
+    "df_busan = pd.read_feather(\"../data/data_for_modeling/df_busan.feather\")\n",
+    "df_incheon = pd.read_feather(\"../data/data_for_modeling/df_incheon.feather\")\n",
+    "df_daegu = pd.read_feather(\"../data/data_for_modeling/df_daegu.feather\")\n",
+    "df_daejeon = pd.read_feather(\"../data/data_for_modeling/df_daejeon.feather\")\n",
+    "df_gwangju = pd.read_feather(\"../data/data_for_modeling/df_gwangju.feather\")"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 3,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "Counter({2: 48534, 1: 3941, 0: 109})"
+      ]
+     },
+     "execution_count": 3,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "Counter(df_seoul['multi_class'])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 4,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "Counter({2: 50069, 1: 2350, 0: 165})"
+      ]
+     },
+     "execution_count": 4,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "Counter(df_busan['multi_class'])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 5,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "Counter({2: 44944, 1: 6658, 0: 982})"
+      ]
+     },
+     "execution_count": 5,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "Counter(df_incheon['multi_class'])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 6,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "Counter({2: 50919, 1: 1610, 0: 55})"
+      ]
+     },
+     "execution_count": 6,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "Counter(df_daegu['multi_class'])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 7,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "Counter({2: 48047, 1: 4227, 0: 310})"
+      ]
+     },
+     "execution_count": 7,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "Counter(df_daejeon['multi_class'])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 8,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "Counter({2: 48405, 1: 4015, 0: 164})"
+      ]
+     },
+     "execution_count": 8,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "Counter(df_gwangju['multi_class'])"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "(52584, 30)"
+      ]
+     },
+     "execution_count": 9,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "df_seoul.shape"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 10,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df_seoul = df_seoul.loc[df_seoul['year'].isin([2018, 2019, 2020, 2021]),:].copy()\n",
+    "df_busan = df_busan.loc[df_busan['year'].isin([2018, 2019, 2020, 2021]),:].copy()\n",
+    "df_incheon = df_incheon.loc[df_incheon['year'].isin([2018, 2019, 2020, 2021]),:].copy()\n",
+    "df_daegu = df_daegu.loc[df_daegu['year'].isin([2018, 2019, 2020, 2021]),:].copy()\n",
+    "df_daejeon = df_daejeon.loc[df_daejeon['year'].isin([2018, 2019, 2020, 2021]),:].copy()\n",
+    "df_gwangju = df_gwangju.loc[df_gwangju['year'].isin([2018, 2019, 2020, 2021]),:].copy()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "cols = [col for col in df_seoul.columns if col != \"multi_class\"] + [\"multi_class\"]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 12,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df_seoul = df_seoul[cols]\n",
+    "df_busan = df_busan[cols]\n",
+    "df_incheon = df_incheon[cols]\n",
+    "df_daegu = df_daegu[cols]\n",
+    "df_daejeon = df_daejeon[cols]\n",
+    "df_gwangju = df_gwangju[cols]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 13,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df_seoul_train = df_seoul.loc[df_seoul['year'].isin([2018, 2019, 2020]),:].copy()\n",
+    "df_seoul_test = df_seoul.loc[df_seoul['year'].isin([2021]),:].copy()\n",
+    "\n",
+    "df_busan_train = df_busan.loc[df_busan['year'].isin([2018, 2019, 2020]),:].copy()\n",
+    "df_busan_test = df_busan.loc[df_busan['year'].isin([2021]),:].copy()\n",
+    "\n",
+    "df_incheon_train = df_incheon.loc[df_incheon['year'].isin([2018, 2019, 2020]),:].copy()\n",
+    "df_incheon_test = df_incheon.loc[df_incheon['year'].isin([2021]),:].copy()\n",
+    "\n",
+    "df_daegu_train = df_daegu.loc[df_daegu['year'].isin([2018, 2019, 2020]),:].copy()\n",
+    "df_daegu_test = df_daegu.loc[df_daegu['year'].isin([2021]),:].copy()\n",
+    "\n",
+    "df_daejeon_train = df_daejeon.loc[df_daejeon['year'].isin([2018, 2019, 2020]),:].copy()\n",
+    "df_daejeon_test = df_daejeon.loc[df_daejeon['year'].isin([2021]),:].copy()\n",
+    "\n",
+    "df_gwangju_train = df_gwangju.loc[df_gwangju['year'].isin([2018, 2019, 2020]),:].copy()\n",
+    "df_gwangju_test = df_gwangju.loc[df_gwangju['year'].isin([2021]),:].copy()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 14,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>temp_C</th>\n",
+       "      <th>precip_mm</th>\n",
+       "      <th>wind_speed</th>\n",
+       "      <th>wind_dir</th>\n",
+       "      <th>hm</th>\n",
+       "      <th>vap_pressure</th>\n",
+       "      <th>dewpoint_C</th>\n",
+       "      <th>loc_pressure</th>\n",
+       "      <th>sea_pressure</th>\n",
+       "      <th>solarRad</th>\n",
+       "      <th>...</th>\n",
+       "      <th>year</th>\n",
+       "      <th>month</th>\n",
+       "      <th>hour</th>\n",
+       "      <th>ground_temp - temp_C</th>\n",
+       "      <th>hour_sin</th>\n",
+       "      <th>hour_cos</th>\n",
+       "      <th>month_sin</th>\n",
+       "      <th>month_cos</th>\n",
+       "      <th>visi</th>\n",
+       "      <th>multi_class</th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "    <tr>\n",
+       "      <th>0</th>\n",
+       "      <td>1.2</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>1.6</td>\n",
+       "      <td>360</td>\n",
+       "      <td>35.0</td>\n",
+       "      <td>2.3</td>\n",
+       "      <td>-12.6</td>\n",
+       "      <td>1015.8</td>\n",
+       "      <td>1024.6</td>\n",
+       "      <td>0.00</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2018</td>\n",
+       "      <td>1</td>\n",
+       "      <td>0</td>\n",
+       "      <td>-5.4</td>\n",
+       "      <td>0.000000</td>\n",
+       "      <td>1.000000e+00</td>\n",
+       "      <td>0.5</td>\n",
+       "      <td>0.866025</td>\n",
+       "      <td>2000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>1</th>\n",
+       "      <td>0.5</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>1.3</td>\n",
+       "      <td>360</td>\n",
+       "      <td>33.0</td>\n",
+       "      <td>2.1</td>\n",
+       "      <td>-13.9</td>\n",
+       "      <td>1015.5</td>\n",
+       "      <td>1024.3</td>\n",
+       "      <td>0.00</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2018</td>\n",
+       "      <td>1</td>\n",
+       "      <td>1</td>\n",
+       "      <td>-5.4</td>\n",
+       "      <td>0.258819</td>\n",
+       "      <td>9.659258e-01</td>\n",
+       "      <td>0.5</td>\n",
+       "      <td>0.866025</td>\n",
+       "      <td>2000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>2</th>\n",
+       "      <td>0.1</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>1.5</td>\n",
+       "      <td>20</td>\n",
+       "      <td>34.0</td>\n",
+       "      <td>2.1</td>\n",
+       "      <td>-13.9</td>\n",
+       "      <td>1015.7</td>\n",
+       "      <td>1024.5</td>\n",
+       "      <td>0.00</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2018</td>\n",
+       "      <td>1</td>\n",
+       "      <td>2</td>\n",
+       "      <td>-5.4</td>\n",
+       "      <td>0.500000</td>\n",
+       "      <td>8.660254e-01</td>\n",
+       "      <td>0.5</td>\n",
+       "      <td>0.866025</td>\n",
+       "      <td>2000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>3</th>\n",
+       "      <td>0.0</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>2.1</td>\n",
+       "      <td>320</td>\n",
+       "      <td>37.0</td>\n",
+       "      <td>2.3</td>\n",
+       "      <td>-12.9</td>\n",
+       "      <td>1015.9</td>\n",
+       "      <td>1024.7</td>\n",
+       "      <td>0.00</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2018</td>\n",
+       "      <td>1</td>\n",
+       "      <td>3</td>\n",
+       "      <td>-5.0</td>\n",
+       "      <td>0.707107</td>\n",
+       "      <td>7.071068e-01</td>\n",
+       "      <td>0.5</td>\n",
+       "      <td>0.866025</td>\n",
+       "      <td>2000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>4</th>\n",
+       "      <td>-0.1</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>2.3</td>\n",
+       "      <td>340</td>\n",
+       "      <td>42.0</td>\n",
+       "      <td>2.5</td>\n",
+       "      <td>-11.5</td>\n",
+       "      <td>1016.0</td>\n",
+       "      <td>1024.9</td>\n",
+       "      <td>0.00</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2018</td>\n",
+       "      <td>1</td>\n",
+       "      <td>4</td>\n",
+       "      <td>-4.3</td>\n",
+       "      <td>0.866025</td>\n",
+       "      <td>5.000000e-01</td>\n",
+       "      <td>0.5</td>\n",
+       "      <td>0.866025</td>\n",
+       "      <td>2000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>5</th>\n",
+       "      <td>-0.1</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>2.8</td>\n",
+       "      <td>50</td>\n",
+       "      <td>43.0</td>\n",
+       "      <td>2.6</td>\n",
+       "      <td>-11.2</td>\n",
+       "      <td>1016.0</td>\n",
+       "      <td>1024.9</td>\n",
+       "      <td>0.00</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2018</td>\n",
+       "      <td>1</td>\n",
+       "      <td>5</td>\n",
+       "      <td>-4.0</td>\n",
+       "      <td>0.965926</td>\n",
+       "      <td>2.588190e-01</td>\n",
+       "      <td>0.5</td>\n",
+       "      <td>0.866025</td>\n",
+       "      <td>2000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>6</th>\n",
+       "      <td>-0.5</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>2.1</td>\n",
+       "      <td>20</td>\n",
+       "      <td>45.0</td>\n",
+       "      <td>2.6</td>\n",
+       "      <td>-11.0</td>\n",
+       "      <td>1016.5</td>\n",
+       "      <td>1025.4</td>\n",
+       "      <td>0.00</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2018</td>\n",
+       "      <td>1</td>\n",
+       "      <td>6</td>\n",
+       "      <td>-4.1</td>\n",
+       "      <td>1.000000</td>\n",
+       "      <td>6.123234e-17</td>\n",
+       "      <td>0.5</td>\n",
+       "      <td>0.866025</td>\n",
+       "      <td>2000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>7</th>\n",
+       "      <td>-0.8</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>2.5</td>\n",
+       "      <td>340</td>\n",
+       "      <td>45.0</td>\n",
+       "      <td>2.6</td>\n",
+       "      <td>-11.2</td>\n",
+       "      <td>1017.1</td>\n",
+       "      <td>1026.0</td>\n",
+       "      <td>0.00</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2018</td>\n",
+       "      <td>1</td>\n",
+       "      <td>7</td>\n",
+       "      <td>-4.5</td>\n",
+       "      <td>0.965926</td>\n",
+       "      <td>-2.588190e-01</td>\n",
+       "      <td>0.5</td>\n",
+       "      <td>0.866025</td>\n",
+       "      <td>2000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>8</th>\n",
+       "      <td>-0.5</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>1.2</td>\n",
+       "      <td>360</td>\n",
+       "      <td>43.0</td>\n",
+       "      <td>2.5</td>\n",
+       "      <td>-11.5</td>\n",
+       "      <td>1017.4</td>\n",
+       "      <td>1026.3</td>\n",
+       "      <td>0.03</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2018</td>\n",
+       "      <td>1</td>\n",
+       "      <td>8</td>\n",
+       "      <td>-4.0</td>\n",
+       "      <td>0.866025</td>\n",
+       "      <td>-5.000000e-01</td>\n",
+       "      <td>0.5</td>\n",
+       "      <td>0.866025</td>\n",
+       "      <td>2000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>9</th>\n",
+       "      <td>1.7</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>2.1</td>\n",
+       "      <td>20</td>\n",
+       "      <td>39.0</td>\n",
+       "      <td>2.7</td>\n",
+       "      <td>-10.8</td>\n",
+       "      <td>1018.1</td>\n",
+       "      <td>1026.9</td>\n",
+       "      <td>0.46</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2018</td>\n",
+       "      <td>1</td>\n",
+       "      <td>9</td>\n",
+       "      <td>2.8</td>\n",
+       "      <td>0.707107</td>\n",
+       "      <td>-7.071068e-01</td>\n",
+       "      <td>0.5</td>\n",
+       "      <td>0.866025</td>\n",
+       "      <td>1953.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "<p>10 rows × 30 columns</p>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "   temp_C  precip_mm  wind_speed wind_dir    hm  vap_pressure  dewpoint_C  \\\n",
+       "0     1.2        0.0         1.6      360  35.0           2.3       -12.6   \n",
+       "1     0.5        0.0         1.3      360  33.0           2.1       -13.9   \n",
+       "2     0.1        0.0         1.5       20  34.0           2.1       -13.9   \n",
+       "3     0.0        0.0         2.1      320  37.0           2.3       -12.9   \n",
+       "4    -0.1        0.0         2.3      340  42.0           2.5       -11.5   \n",
+       "5    -0.1        0.0         2.8       50  43.0           2.6       -11.2   \n",
+       "6    -0.5        0.0         2.1       20  45.0           2.6       -11.0   \n",
+       "7    -0.8        0.0         2.5      340  45.0           2.6       -11.2   \n",
+       "8    -0.5        0.0         1.2      360  43.0           2.5       -11.5   \n",
+       "9     1.7        0.0         2.1       20  39.0           2.7       -10.8   \n",
+       "\n",
+       "   loc_pressure  sea_pressure  solarRad  ...  year month hour  \\\n",
+       "0        1015.8        1024.6      0.00  ...  2018     1    0   \n",
+       "1        1015.5        1024.3      0.00  ...  2018     1    1   \n",
+       "2        1015.7        1024.5      0.00  ...  2018     1    2   \n",
+       "3        1015.9        1024.7      0.00  ...  2018     1    3   \n",
+       "4        1016.0        1024.9      0.00  ...  2018     1    4   \n",
+       "5        1016.0        1024.9      0.00  ...  2018     1    5   \n",
+       "6        1016.5        1025.4      0.00  ...  2018     1    6   \n",
+       "7        1017.1        1026.0      0.00  ...  2018     1    7   \n",
+       "8        1017.4        1026.3      0.03  ...  2018     1    8   \n",
+       "9        1018.1        1026.9      0.46  ...  2018     1    9   \n",
+       "\n",
+       "   ground_temp - temp_C  hour_sin      hour_cos  month_sin  month_cos    visi  \\\n",
+       "0                  -5.4  0.000000  1.000000e+00        0.5   0.866025  2000.0   \n",
+       "1                  -5.4  0.258819  9.659258e-01        0.5   0.866025  2000.0   \n",
+       "2                  -5.4  0.500000  8.660254e-01        0.5   0.866025  2000.0   \n",
+       "3                  -5.0  0.707107  7.071068e-01        0.5   0.866025  2000.0   \n",
+       "4                  -4.3  0.866025  5.000000e-01        0.5   0.866025  2000.0   \n",
+       "5                  -4.0  0.965926  2.588190e-01        0.5   0.866025  2000.0   \n",
+       "6                  -4.1  1.000000  6.123234e-17        0.5   0.866025  2000.0   \n",
+       "7                  -4.5  0.965926 -2.588190e-01        0.5   0.866025  2000.0   \n",
+       "8                  -4.0  0.866025 -5.000000e-01        0.5   0.866025  2000.0   \n",
+       "9                   2.8  0.707107 -7.071068e-01        0.5   0.866025  1953.0   \n",
+       "\n",
+       "   multi_class  \n",
+       "0            2  \n",
+       "1            2  \n",
+       "2            2  \n",
+       "3            2  \n",
+       "4            2  \n",
+       "5            2  \n",
+       "6            2  \n",
+       "7            2  \n",
+       "8            2  \n",
+       "9            2  \n",
+       "\n",
+       "[10 rows x 30 columns]"
+      ]
+     },
+     "execution_count": 14,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "df_busan_train.head(10)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 15,
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/html": [
+       "<div>\n",
+       "<style scoped>\n",
+       "    .dataframe tbody tr th:only-of-type {\n",
+       "        vertical-align: middle;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe tbody tr th {\n",
+       "        vertical-align: top;\n",
+       "    }\n",
+       "\n",
+       "    .dataframe thead th {\n",
+       "        text-align: right;\n",
+       "    }\n",
+       "</style>\n",
+       "<table border=\"1\" class=\"dataframe\">\n",
+       "  <thead>\n",
+       "    <tr style=\"text-align: right;\">\n",
+       "      <th></th>\n",
+       "      <th>temp_C</th>\n",
+       "      <th>precip_mm</th>\n",
+       "      <th>wind_speed</th>\n",
+       "      <th>wind_dir</th>\n",
+       "      <th>hm</th>\n",
+       "      <th>vap_pressure</th>\n",
+       "      <th>dewpoint_C</th>\n",
+       "      <th>loc_pressure</th>\n",
+       "      <th>sea_pressure</th>\n",
+       "      <th>solarRad</th>\n",
+       "      <th>...</th>\n",
+       "      <th>year</th>\n",
+       "      <th>month</th>\n",
+       "      <th>hour</th>\n",
+       "      <th>ground_temp - temp_C</th>\n",
+       "      <th>hour_sin</th>\n",
+       "      <th>hour_cos</th>\n",
+       "      <th>month_sin</th>\n",
+       "      <th>month_cos</th>\n",
+       "      <th>visi</th>\n",
+       "      <th>multi_class</th>\n",
+       "    </tr>\n",
+       "  </thead>\n",
+       "  <tbody>\n",
+       "    <tr>\n",
+       "      <th>26294</th>\n",
+       "      <td>0.1</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>6.3</td>\n",
+       "      <td>270</td>\n",
+       "      <td>37.0</td>\n",
+       "      <td>2.3</td>\n",
+       "      <td>-12.9</td>\n",
+       "      <td>1013.3</td>\n",
+       "      <td>1022.1</td>\n",
+       "      <td>2.07</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2020</td>\n",
+       "      <td>12</td>\n",
+       "      <td>14</td>\n",
+       "      <td>5.8</td>\n",
+       "      <td>-0.500000</td>\n",
+       "      <td>-8.660254e-01</td>\n",
+       "      <td>-2.449294e-16</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>5000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>26295</th>\n",
+       "      <td>1.2</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>5.9</td>\n",
+       "      <td>270</td>\n",
+       "      <td>35.0</td>\n",
+       "      <td>2.3</td>\n",
+       "      <td>-12.6</td>\n",
+       "      <td>1013.2</td>\n",
+       "      <td>1022.0</td>\n",
+       "      <td>1.71</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2020</td>\n",
+       "      <td>12</td>\n",
+       "      <td>15</td>\n",
+       "      <td>5.6</td>\n",
+       "      <td>-0.707107</td>\n",
+       "      <td>-7.071068e-01</td>\n",
+       "      <td>-2.449294e-16</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>5000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>26296</th>\n",
+       "      <td>1.6</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>3.6</td>\n",
+       "      <td>290</td>\n",
+       "      <td>34.0</td>\n",
+       "      <td>2.3</td>\n",
+       "      <td>-12.6</td>\n",
+       "      <td>1012.8</td>\n",
+       "      <td>1021.6</td>\n",
+       "      <td>1.14</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2020</td>\n",
+       "      <td>12</td>\n",
+       "      <td>16</td>\n",
+       "      <td>1.4</td>\n",
+       "      <td>-0.866025</td>\n",
+       "      <td>-5.000000e-01</td>\n",
+       "      <td>-2.449294e-16</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>5000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>26297</th>\n",
+       "      <td>1.2</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>3.8</td>\n",
+       "      <td>250</td>\n",
+       "      <td>38.0</td>\n",
+       "      <td>2.5</td>\n",
+       "      <td>-11.5</td>\n",
+       "      <td>1012.8</td>\n",
+       "      <td>1021.6</td>\n",
+       "      <td>0.48</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2020</td>\n",
+       "      <td>12</td>\n",
+       "      <td>17</td>\n",
+       "      <td>-0.4</td>\n",
+       "      <td>-0.965926</td>\n",
+       "      <td>-2.588190e-01</td>\n",
+       "      <td>-2.449294e-16</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>5000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>26298</th>\n",
+       "      <td>0.9</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>3.8</td>\n",
+       "      <td>270</td>\n",
+       "      <td>40.0</td>\n",
+       "      <td>2.6</td>\n",
+       "      <td>-11.2</td>\n",
+       "      <td>1013.1</td>\n",
+       "      <td>1021.9</td>\n",
+       "      <td>0.02</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2020</td>\n",
+       "      <td>12</td>\n",
+       "      <td>18</td>\n",
+       "      <td>-0.8</td>\n",
+       "      <td>-1.000000</td>\n",
+       "      <td>-1.836970e-16</td>\n",
+       "      <td>-2.449294e-16</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>5000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>26299</th>\n",
+       "      <td>0.6</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>6.2</td>\n",
+       "      <td>270</td>\n",
+       "      <td>41.0</td>\n",
+       "      <td>2.6</td>\n",
+       "      <td>-11.1</td>\n",
+       "      <td>1014.0</td>\n",
+       "      <td>1022.8</td>\n",
+       "      <td>0.00</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2020</td>\n",
+       "      <td>12</td>\n",
+       "      <td>19</td>\n",
+       "      <td>-1.1</td>\n",
+       "      <td>-0.965926</td>\n",
+       "      <td>2.588190e-01</td>\n",
+       "      <td>-2.449294e-16</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>5000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>26300</th>\n",
+       "      <td>0.1</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>6.0</td>\n",
+       "      <td>270</td>\n",
+       "      <td>44.0</td>\n",
+       "      <td>2.7</td>\n",
+       "      <td>-10.7</td>\n",
+       "      <td>1014.8</td>\n",
+       "      <td>1023.6</td>\n",
+       "      <td>0.00</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2020</td>\n",
+       "      <td>12</td>\n",
+       "      <td>20</td>\n",
+       "      <td>-0.9</td>\n",
+       "      <td>-0.866025</td>\n",
+       "      <td>5.000000e-01</td>\n",
+       "      <td>-2.449294e-16</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>5000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>26301</th>\n",
+       "      <td>-0.2</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>5.0</td>\n",
+       "      <td>290</td>\n",
+       "      <td>48.0</td>\n",
+       "      <td>2.9</td>\n",
+       "      <td>-9.9</td>\n",
+       "      <td>1014.6</td>\n",
+       "      <td>1023.4</td>\n",
+       "      <td>0.00</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2020</td>\n",
+       "      <td>12</td>\n",
+       "      <td>21</td>\n",
+       "      <td>-0.8</td>\n",
+       "      <td>-0.707107</td>\n",
+       "      <td>7.071068e-01</td>\n",
+       "      <td>-2.449294e-16</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>5000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>26302</th>\n",
+       "      <td>-0.7</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>2.7</td>\n",
+       "      <td>270</td>\n",
+       "      <td>51.0</td>\n",
+       "      <td>3.0</td>\n",
+       "      <td>-9.6</td>\n",
+       "      <td>1014.8</td>\n",
+       "      <td>1023.6</td>\n",
+       "      <td>0.00</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2020</td>\n",
+       "      <td>12</td>\n",
+       "      <td>22</td>\n",
+       "      <td>-0.6</td>\n",
+       "      <td>-0.500000</td>\n",
+       "      <td>8.660254e-01</td>\n",
+       "      <td>-2.449294e-16</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>5000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "    <tr>\n",
+       "      <th>26303</th>\n",
+       "      <td>-0.7</td>\n",
+       "      <td>0.0</td>\n",
+       "      <td>3.8</td>\n",
+       "      <td>250</td>\n",
+       "      <td>55.0</td>\n",
+       "      <td>3.2</td>\n",
+       "      <td>-8.6</td>\n",
+       "      <td>1015.1</td>\n",
+       "      <td>1024.0</td>\n",
+       "      <td>0.00</td>\n",
+       "      <td>...</td>\n",
+       "      <td>2020</td>\n",
+       "      <td>12</td>\n",
+       "      <td>23</td>\n",
+       "      <td>-0.6</td>\n",
+       "      <td>-0.258819</td>\n",
+       "      <td>9.659258e-01</td>\n",
+       "      <td>-2.449294e-16</td>\n",
+       "      <td>1.0</td>\n",
+       "      <td>5000.0</td>\n",
+       "      <td>2</td>\n",
+       "    </tr>\n",
+       "  </tbody>\n",
+       "</table>\n",
+       "<p>10 rows × 30 columns</p>\n",
+       "</div>"
+      ],
+      "text/plain": [
+       "       temp_C  precip_mm  wind_speed wind_dir    hm  vap_pressure  dewpoint_C  \\\n",
+       "26294     0.1        0.0         6.3      270  37.0           2.3       -12.9   \n",
+       "26295     1.2        0.0         5.9      270  35.0           2.3       -12.6   \n",
+       "26296     1.6        0.0         3.6      290  34.0           2.3       -12.6   \n",
+       "26297     1.2        0.0         3.8      250  38.0           2.5       -11.5   \n",
+       "26298     0.9        0.0         3.8      270  40.0           2.6       -11.2   \n",
+       "26299     0.6        0.0         6.2      270  41.0           2.6       -11.1   \n",
+       "26300     0.1        0.0         6.0      270  44.0           2.7       -10.7   \n",
+       "26301    -0.2        0.0         5.0      290  48.0           2.9        -9.9   \n",
+       "26302    -0.7        0.0         2.7      270  51.0           3.0        -9.6   \n",
+       "26303    -0.7        0.0         3.8      250  55.0           3.2        -8.6   \n",
+       "\n",
+       "       loc_pressure  sea_pressure  solarRad  ...  year month hour  \\\n",
+       "26294        1013.3        1022.1      2.07  ...  2020    12   14   \n",
+       "26295        1013.2        1022.0      1.71  ...  2020    12   15   \n",
+       "26296        1012.8        1021.6      1.14  ...  2020    12   16   \n",
+       "26297        1012.8        1021.6      0.48  ...  2020    12   17   \n",
+       "26298        1013.1        1021.9      0.02  ...  2020    12   18   \n",
+       "26299        1014.0        1022.8      0.00  ...  2020    12   19   \n",
+       "26300        1014.8        1023.6      0.00  ...  2020    12   20   \n",
+       "26301        1014.6        1023.4      0.00  ...  2020    12   21   \n",
+       "26302        1014.8        1023.6      0.00  ...  2020    12   22   \n",
+       "26303        1015.1        1024.0      0.00  ...  2020    12   23   \n",
+       "\n",
+       "       ground_temp - temp_C  hour_sin      hour_cos     month_sin  month_cos  \\\n",
+       "26294                   5.8 -0.500000 -8.660254e-01 -2.449294e-16        1.0   \n",
+       "26295                   5.6 -0.707107 -7.071068e-01 -2.449294e-16        1.0   \n",
+       "26296                   1.4 -0.866025 -5.000000e-01 -2.449294e-16        1.0   \n",
+       "26297                  -0.4 -0.965926 -2.588190e-01 -2.449294e-16        1.0   \n",
+       "26298                  -0.8 -1.000000 -1.836970e-16 -2.449294e-16        1.0   \n",
+       "26299                  -1.1 -0.965926  2.588190e-01 -2.449294e-16        1.0   \n",
+       "26300                  -0.9 -0.866025  5.000000e-01 -2.449294e-16        1.0   \n",
+       "26301                  -0.8 -0.707107  7.071068e-01 -2.449294e-16        1.0   \n",
+       "26302                  -0.6 -0.500000  8.660254e-01 -2.449294e-16        1.0   \n",
+       "26303                  -0.6 -0.258819  9.659258e-01 -2.449294e-16        1.0   \n",
+       "\n",
+       "         visi  multi_class  \n",
+       "26294  5000.0            2  \n",
+       "26295  5000.0            2  \n",
+       "26296  5000.0            2  \n",
+       "26297  5000.0            2  \n",
+       "26298  5000.0            2  \n",
+       "26299  5000.0            2  \n",
+       "26300  5000.0            2  \n",
+       "26301  5000.0            2  \n",
+       "26302  5000.0            2  \n",
+       "26303  5000.0            2  \n",
+       "\n",
+       "[10 rows x 30 columns]"
+      ]
+     },
+     "execution_count": 15,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "df_busan_train.tail(10)"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 16,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "<class 'pandas.core.frame.DataFrame'>\n",
+      "Index: 26304 entries, 0 to 26303\n",
+      "Data columns (total 30 columns):\n",
+      " #   Column                Non-Null Count  Dtype   \n",
+      "---  ------                --------------  -----   \n",
+      " 0   temp_C                26304 non-null  float64 \n",
+      " 1   precip_mm             26304 non-null  float64 \n",
+      " 2   wind_speed            26304 non-null  float64 \n",
+      " 3   wind_dir              26304 non-null  category\n",
+      " 4   hm                    26304 non-null  float64 \n",
+      " 5   vap_pressure          26304 non-null  float64 \n",
+      " 6   dewpoint_C            26304 non-null  float64 \n",
+      " 7   loc_pressure          26304 non-null  float64 \n",
+      " 8   sea_pressure          26304 non-null  float64 \n",
+      " 9   solarRad              26304 non-null  float64 \n",
+      " 10  snow_cm               26304 non-null  float64 \n",
+      " 11  cloudcover            26304 non-null  category\n",
+      " 12  lm_cloudcover         26304 non-null  category\n",
+      " 13  low_cloudbase         26304 non-null  float64 \n",
+      " 14  groundtemp            26304 non-null  float64 \n",
+      " 15  O3                    26304 non-null  float64 \n",
+      " 16  NO2                   26304 non-null  float64 \n",
+      " 17  PM10                  26304 non-null  float64 \n",
+      " 18  PM25                  26304 non-null  float64 \n",
+      " 19  binary_class          26304 non-null  int64   \n",
+      " 20  year                  26304 non-null  int64   \n",
+      " 21  month                 26304 non-null  int64   \n",
+      " 22  hour                  26304 non-null  int64   \n",
+      " 23  ground_temp - temp_C  26304 non-null  float64 \n",
+      " 24  hour_sin              26304 non-null  float64 \n",
+      " 25  hour_cos              26304 non-null  float64 \n",
+      " 26  month_sin             26304 non-null  float64 \n",
+      " 27  month_cos             26304 non-null  float64 \n",
+      " 28  visi                  26304 non-null  float64 \n",
+      " 29  multi_class           26304 non-null  int64   \n",
+      "dtypes: category(3), float64(22), int64(5)\n",
+      "memory usage: 5.7 MB\n"
+     ]
+    }
+   ],
+   "source": [
+    "df_busan_train.info()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 17,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "df_seoul_train.to_csv(\"../data/data_for_modeling/seoul_train.csv\")\n",
+    "df_seoul_test.to_csv(\"../data/data_for_modeling/seoul_test.csv\")\n",
+    "\n",
+    "df_busan_train.to_csv(\"../data/data_for_modeling/busan_train.csv\")\n",
+    "df_busan_test.to_csv(\"../data/data_for_modeling/busan_test.csv\")\n",
+    "\n",
+    "df_incheon_train.to_csv(\"../data/data_for_modeling/incheon_train.csv\")\n",
+    "df_incheon_test.to_csv(\"../data/data_for_modeling/incheon_test.csv\")\n",
+    "\n",
+    "df_daegu_train.to_csv(\"../data/data_for_modeling/daegu_train.csv\")\n",
+    "df_daegu_test.to_csv(\"../data/data_for_modeling/daegu_test.csv\")\n",
+    "\n",
+    "df_daejeon_train.to_csv(\"../data/data_for_modeling/daejeon_train.csv\")\n",
+    "df_daejeon_test.to_csv(\"../data/data_for_modeling/daejeon_test.csv\")\n",
+    "\n",
+    "df_gwangju_train.to_csv(\"../data/data_for_modeling/gwangju_train.csv\")\n",
+    "df_gwangju_test.to_csv(\"../data/data_for_modeling/gwangju_test.csv\")\n",
+    "\n",
+    "df_seoul_train = pd.read_csv(\"../data/data_for_modeling/seoul_train.csv\")\n",
+    "df_seoul_test = pd.read_csv(\"../data/data_for_modeling/seoul_test.csv\")\n"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 18,
+   "metadata": {},
+   "outputs": [
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "Counter({2: 8266, 1: 481, 0: 13})\n",
+      "Counter({2: 23686, 1: 2579, 0: 39})\n",
+      "Counter({2: 8455, 1: 281, 0: 24})\n",
+      "Counter({2: 24694, 1: 1516, 0: 94})\n",
+      "Counter({2: 7373, 1: 1205, 0: 182})\n",
+      "Counter({2: 21893, 1: 3892, 0: 519})\n",
+      "Counter({2: 8631, 1: 128, 0: 1})\n",
+      "Counter({2: 25149, 1: 1107, 0: 48})\n",
+      "Counter({2: 8089, 1: 618, 0: 53})\n",
+      "Counter({2: 23471, 1: 2660, 0: 173})\n",
+      "Counter({2: 8087, 1: 643, 0: 30})\n",
+      "Counter({2: 23798, 1: 2411, 0: 95})\n"
+     ]
+    }
+   ],
+   "source": [
+    "print(Counter(df_seoul_test['multi_class']))\n",
+    "print(Counter(df_seoul_train['multi_class']))\n",
+    "\n",
+    "print(Counter(df_busan_test['multi_class']))\n",
+    "print(Counter(df_busan_train['multi_class']))\n",
+    "\n",
+    "print(Counter(df_incheon_test['multi_class']))\n",
+    "print(Counter(df_incheon_train['multi_class']))\n",
+    "\n",
+    "print(Counter(df_daegu_test['multi_class']))\n",
+    "print(Counter(df_daegu_train['multi_class']))\n",
+    "\n",
+    "print(Counter(df_daejeon_test['multi_class']))\n",
+    "print(Counter(df_daejeon_train['multi_class']))\n",
+    "\n",
+    "print(Counter(df_gwangju_test['multi_class']))\n",
+    "print(Counter(df_gwangju_train['multi_class']))"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.8.10"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 2
+}

Analysis_code/model_result/best_sample/ensemble_best_sample.csv

@@ -0,0 +1,157 @@
+model,CSI,MCC,Accuracy,region,data_sample
+deepgbm+ft_transformer,0.6992424754332419,0.8049778134563997,0.9617050719032529,seoul,best
+deepgbm+resnet_like,0.7090721222546449,0.8134996041947694,0.9639575067994112,seoul,best
+deepgbm+XGBoost,0.6692268948642789,0.7828377071996622,0.9579950262411524,seoul,best
+deepgbm+LightGBM,0.6935149238495999,0.801572479004747,0.9620935948299524,seoul,best
+ft_transformer+resnet_like,0.6252731559795573,0.7509360530096033,0.9488043848924154,seoul,best
+ft_transformer+XGBoost,0.6142839547044981,0.7403116223696319,0.9470938235563208,seoul,best
+ft_transformer+LightGBM,0.6085686211212852,0.7359331482423171,0.9466367866856302,seoul,best
+resnet_like+XGBoost,0.6194270544060083,0.7446567256109252,0.9478222130731675,seoul,best
+resnet_like+LightGBM,0.6136847718241535,0.741095701007121,0.9457727208457053,seoul,best
+XGBoost+LightGBM,0.5861106806341969,0.7149062718037028,0.9421973118413721,seoul,best
+deepgbm+ft_transformer+resnet_like,0.6821685259093041,0.7946632359980877,0.9596256373148356,seoul,best
+deepgbm+ft_transformer+XGBoost,0.6730333196442752,0.7867573605932577,0.9584481123836616,seoul,best
+deepgbm+ft_transformer+LightGBM,0.6764764169794865,0.7902017631994056,0.9590548610591277,seoul,best
+deepgbm+resnet_like+XGBoost,0.6784375221650579,0.7915054245115618,0.9591382422170655,seoul,best
+deepgbm+resnet_like+LightGBM,0.6775625748710835,0.7913206958685782,0.9590235671332685,seoul,best
+deepgbm+XGBoost+LightGBM,0.6325378315490732,0.7547274504003308,0.9511999817018905,seoul,best
+ft_transformer+resnet_like+XGBoost,0.6332314687362993,0.7561343704178952,0.9500612362868145,seoul,best
+ft_transformer+resnet_like+LightGBM,0.6306548804840734,0.7545886604473321,0.9496810306826026,seoul,best
+ft_transformer+XGBoost+LightGBM,0.6052879958628536,0.7319675532436248,0.9455772637672482,seoul,best
+resnet_like+XGBoost+LightGBM,0.6103490502864374,0.7369455626993492,0.9460374196338716,seoul,best
+deepgbm+ft_transformer+resnet_like+XGBoost,0.6729967545904317,0.7875845379790564,0.9581108449567916,seoul,best
+deepgbm+ft_transformer+resnet_like+LightGBM,0.6719561859986568,0.7873723840805852,0.9580337017907196,seoul,best
+deepgbm+ft_transformer+XGBoost+LightGBM,0.6531335492225346,0.7716405213564462,0.9549966314843926,seoul,best
+deepgbm+resnet_like+XGBoost+LightGBM,0.6495283344540792,0.769203226760809,0.9539740166845488,seoul,best
+ft_transformer+resnet_like+XGBoost+LightGBM,0.6233313312481723,0.7475113353931637,0.9482768587136429,seoul,best
+deepgbm+ft_transformer+resnet_like+XGBoost+LightGBM,0.6569118107203898,0.7752226196801869,0.955377564854322,seoul,best
+deepgbm+ft_transformer,0.6231286091598959,0.762319461740938,0.9692817368232819,busan,best
+deepgbm+resnet_like,0.693965736363066,0.8126142282014271,0.9774987316083207,busan,best
+deepgbm+XGBoost,0.5949465703841524,0.7372757652859034,0.9695935324500337,busan,best
+deepgbm+LightGBM,0.6134672160989406,0.7540785415035627,0.969397659505452,busan,best
+ft_transformer+resnet_like,0.5993289393140034,0.7415869283802814,0.9687921584283586,busan,best
+ft_transformer+XGBoost,0.5275703861594696,0.6836131837976293,0.9627103242924039,busan,best
+ft_transformer+LightGBM,0.5340117437566735,0.690570910687501,0.9611092397135513,busan,best
+resnet_like+XGBoost,0.5476348496496576,0.6975679320922632,0.9653371426670326,busan,best
+resnet_like+LightGBM,0.5541563454462936,0.7052881357398278,0.9637749415708096,busan,best
+XGBoost+LightGBM,0.4789290487253062,0.6395338303931094,0.9572717310843294,busan,best
+deepgbm+ft_transformer+resnet_like,0.6446887106206897,0.777619884223958,0.9728593291247681,busan,best
+deepgbm+ft_transformer+XGBoost,0.6102029002552042,0.7511249566662123,0.9699326712744633,busan,best
+deepgbm+ft_transformer+LightGBM,0.6162985577163776,0.7570233197218496,0.9692066729878318,busan,best
+deepgbm+resnet_like+XGBoost,0.6261344637257955,0.7612451856135231,0.9726364248821019,busan,best
+deepgbm+resnet_like+LightGBM,0.6357139958459742,0.770523881133966,0.9720613859986192,busan,best
+deepgbm+XGBoost+LightGBM,0.550954042564896,0.7024853673988001,0.9645722608977718,busan,best
+ft_transformer+resnet_like+XGBoost,0.5708235070191732,0.7183051671189992,0.9673502466086118,busan,best
+ft_transformer+resnet_like+LightGBM,0.5747801644059322,0.723481589964096,0.9662444585838926,busan,best
+ft_transformer+XGBoost+LightGBM,0.5188547215134367,0.6758426217880089,0.9612265139606259,busan,best
+resnet_like+XGBoost+LightGBM,0.5283032797443571,0.6828500944597963,0.9625962730077933,busan,best
+deepgbm+ft_transformer+resnet_like+XGBoost,0.62252465197324,0.7595480882291313,0.9716073641573305,busan,best
+deepgbm+ft_transformer+resnet_like+LightGBM,0.6273921955959187,0.7650978772099609,0.9711861957398673,busan,best
+deepgbm+ft_transformer+XGBoost+LightGBM,0.5803171388664561,0.7273427892946641,0.9671579085260872,busan,best
+deepgbm+resnet_like+XGBoost+LightGBM,0.5883015315947713,0.7322310098800981,0.9685279794728481,busan,best
+ft_transformer+resnet_like+XGBoost+LightGBM,0.5527408442450511,0.7040063830266932,0.9648394548826841,busan,best
+deepgbm+ft_transformer+resnet_like+XGBoost+LightGBM,0.5949050119452819,0.7385090163673939,0.9689076652444045,busan,best
+deepgbm+ft_transformer,0.5873884001633557,0.7086005084355499,0.9163964576523526,incheon,best
+deepgbm+resnet_like,0.5938343436639008,0.7129218344958757,0.9139223661119011,incheon,best
+deepgbm+XGBoost,0.5919031180535835,0.7111840783628853,0.9141871688665986,incheon,best
+deepgbm+LightGBM,0.5936054700869063,0.71280494431202,0.9151763064434298,incheon,best
+ft_transformer+resnet_like,0.5967079690105518,0.7167701525416347,0.9161293676339713,incheon,best
+ft_transformer+XGBoost,0.5958609493419124,0.7170391565071776,0.9169653625105005,incheon,best
+ft_transformer+LightGBM,0.5970916463252486,0.7186247751024871,0.9177277490830152,incheon,best
+resnet_like+XGBoost,0.6048691059122082,0.7230658352305586,0.9150990593108267,incheon,best
+resnet_like+LightGBM,0.6006642978391444,0.7204725929547785,0.913995974415916,incheon,best
+XGBoost+LightGBM,0.5923037998052801,0.7130252479770401,0.9127417221847942,incheon,best
+deepgbm+ft_transformer+resnet_like,0.5991788818331828,0.7177110572369007,0.9171559331619964,incheon,best
+deepgbm+ft_transformer+XGBoost,0.5957796816002722,0.7161817507572641,0.9171555172958721,incheon,best
+deepgbm+ft_transformer+LightGBM,0.5956266359998414,0.7160886468480842,0.9173455681147126,incheon,best
+deepgbm+resnet_like+XGBoost,0.606730149793507,0.7239728449518917,0.9167729204614451,incheon,best
+deepgbm+resnet_like+LightGBM,0.6063493089884,0.7235778913430906,0.9167717768296031,incheon,best
+deepgbm+XGBoost+LightGBM,0.6018765631426876,0.7202419676609231,0.9160501451372774,incheon,best
+ft_transformer+resnet_like+XGBoost,0.6044191487414033,0.7234603140558615,0.9175339554690555,incheon,best
+ft_transformer+resnet_like+LightGBM,0.6061367487233379,0.7248883683993704,0.9179906804401528,incheon,best
+ft_transformer+XGBoost+LightGBM,0.6013258447351273,0.721556226775587,0.9174587876670742,incheon,best
+resnet_like+XGBoost+LightGBM,0.6028836532105156,0.7216380012607432,0.915023475642721,incheon,best
+deepgbm+ft_transformer+resnet_like+XGBoost,0.6058688939038325,0.7235044499077539,0.9180666799743826,incheon,best
+deepgbm+ft_transformer+resnet_like+LightGBM,0.6053266964324343,0.7232391120300768,0.9180279004582844,incheon,best
+deepgbm+ft_transformer+XGBoost+LightGBM,0.6030917448851759,0.7220222485555309,0.9181051475908876,incheon,best
+deepgbm+resnet_like+XGBoost+LightGBM,0.6054312862809185,0.7231150429381622,0.9164303507414893,incheon,best
+ft_transformer+resnet_like+XGBoost+LightGBM,0.6046920079336132,0.7233826726564859,0.9171152822483387,incheon,best
+deepgbm+ft_transformer+resnet_like+XGBoost+LightGBM,0.6109420059115421,0.72786127635734,0.9189396869359815,incheon,best
+deepgbm+ft_transformer,0.6446005403603131,0.7684243711794304,0.9801833553742378,daegu,best
+deepgbm+resnet_like,0.6137498386585009,0.7496378336201855,0.9772962048057489,daegu,best
+deepgbm+XGBoost,0.5974990513708366,0.7398559419115336,0.9765007568763463,daegu,best
+deepgbm+LightGBM,0.6115123902876293,0.7497436182930696,0.9776771381756785,daegu,best
+ft_transformer+resnet_like,0.5469112144188149,0.6944690161411611,0.9743305595062838,daegu,best
+ft_transformer+XGBoost,0.5161941376039343,0.6722802482596641,0.9719755096439354,daegu,best
+ft_transformer+LightGBM,0.5073005220098658,0.6635248688295593,0.9720125217290049,daegu,best
+resnet_like+XGBoost,0.5029516970497304,0.6645749569024574,0.9683687027472115,daegu,best
+resnet_like+LightGBM,0.4951131290074195,0.6589965524803283,0.968139976378804,daegu,best
+XGBoost+LightGBM,0.4549973810817563,0.6238297325596118,0.9647562816578087,daegu,best
+deepgbm+ft_transformer+resnet_like,0.6183060943020874,0.7512947555661068,0.978549833237684,daegu,best
+deepgbm+ft_transformer+XGBoost,0.6017176745611856,0.7392748101693137,0.9775999950096065,daegu,best
+deepgbm+ft_transformer+LightGBM,0.6004996462619208,0.7388281525299276,0.9778661493292079,daegu,best
+deepgbm+resnet_like+XGBoost,0.5739664154870325,0.7226992953385828,0.9747144039390839,daegu,best
+deepgbm+resnet_like+LightGBM,0.5760590538958189,0.7240729842389874,0.975132349394083,daegu,best
+deepgbm+XGBoost+LightGBM,0.5273980786589871,0.6866104535247874,0.9714085801498781,daegu,best
+ft_transformer+resnet_like+XGBoost,0.5287822236272242,0.6819314770838893,0.9723570668130517,daegu,best
+ft_transformer+resnet_like+LightGBM,0.520853018041326,0.675931210971238,0.972014705026158,daegu,best
+ft_transformer+XGBoost+LightGBM,0.48437190164347993,0.6467498445011408,0.9690512430238457,daegu,best
+resnet_like+XGBoost+LightGBM,0.4803327637605557,0.6466009408894583,0.9668476723973018,daegu,best
+deepgbm+ft_transformer+resnet_like+XGBoost,0.5787636584539054,0.7225361676622583,0.9760432001730003,daegu,best
+deepgbm+ft_transformer+resnet_like+LightGBM,0.5827514275296815,0.7252407488954487,0.9763842103949898,daegu,best
+deepgbm+ft_transformer+XGBoost+LightGBM,0.5563231124331217,0.7066219754268251,0.9745609493391888,daegu,best
+deepgbm+resnet_like+XGBoost+LightGBM,0.5497033929377105,0.7036672133164642,0.9729662067187331,daegu,best
+ft_transformer+resnet_like+XGBoost+LightGBM,0.5057939271151667,0.6644447331953721,0.9703813908226664,daegu,best
+deepgbm+ft_transformer+resnet_like+XGBoost+LightGBM,0.5538086053234044,0.7036101925279703,0.9742201470502616,daegu,best
+deepgbm+ft_transformer,0.6798477338109503,0.7890627213525137,0.9581849730934616,daejeon,best
+deepgbm+resnet_like,0.665680366173557,0.7799054500657331,0.9548432808510284,daejeon,best
+deepgbm+XGBoost,0.6574065703012586,0.7732790182043394,0.9541580374445858,daejeon,best
+deepgbm+LightGBM,0.6615432493953055,0.7763462848232718,0.9551088113714433,daejeon,best
+ft_transformer+resnet_like,0.5874777029243061,0.7124933721407519,0.9448812078415716,daejeon,best
+ft_transformer+XGBoost,0.5819133878684225,0.7079795420327318,0.9433591378263508,daejeon,best
+ft_transformer+LightGBM,0.5805334443959108,0.7070987268766625,0.9437394473970939,daejeon,best
+resnet_like+XGBoost,0.5621088257073251,0.6925993749297947,0.9389891126248638,daejeon,best
+resnet_like+LightGBM,0.5552708987061287,0.6862528661139199,0.9386090109871829,daejeon,best
+XGBoost+LightGBM,0.5288440675809284,0.6634738950206458,0.9327136928080112,daejeon,best
+deepgbm+ft_transformer+resnet_like,0.6605998838783642,0.7743701608802702,0.9553335870116691,daejeon,best
+deepgbm+ft_transformer+XGBoost,0.6495239084733958,0.7651993867637442,0.9536631567565769,daejeon,best
+deepgbm+ft_transformer+LightGBM,0.6556854096497969,0.7699591127202252,0.9550316682053713,daejeon,best
+deepgbm+resnet_like+XGBoost,0.646377396487214,0.7644362437422076,0.9526370070946761,daejeon,best
+deepgbm+resnet_like+LightGBM,0.6441717942778256,0.7623630196443353,0.9527517861450042,daejeon,best
+deepgbm+XGBoost+LightGBM,0.6104936580055047,0.7355161749128437,0.9470099225657277,daejeon,best
+ft_transformer+resnet_like+XGBoost,0.5926724219238438,0.7169506517948077,0.9453380367792,daejeon,best
+ft_transformer+resnet_like+LightGBM,0.5868504053718506,0.7120674745305696,0.9448050003742795,daejeon,best
+ft_transformer+XGBoost+LightGBM,0.5697749480416849,0.6978540261423789,0.9408133093794446,daejeon,best
+resnet_like+XGBoost+LightGBM,0.5568421888456356,0.6879639430352676,0.9383421289018639,daejeon,best
+deepgbm+ft_transformer+resnet_like+XGBoost,0.6464484806055241,0.7628617034260374,0.953282847185834,daejeon,best
+deepgbm+ft_transformer+resnet_like+LightGBM,0.6432761028976647,0.7603027833589299,0.9530166928662326,daejeon,best
+deepgbm+ft_transformer+XGBoost+LightGBM,0.6268590959555621,0.7470252829879999,0.9501275669336527,daejeon,best
+deepgbm+resnet_like+XGBoost+LightGBM,0.617990326312608,0.7412687534166494,0.9484927971987257,daejeon,best
+ft_transformer+resnet_like+XGBoost+LightGBM,0.5778772266115083,0.7046854387067157,0.9426776372150277,daejeon,best
+deepgbm+ft_transformer+resnet_like+XGBoost+LightGBM,0.6251709503819354,0.7454103687111592,0.9500901389824588,daejeon,best
+deepgbm+ft_transformer,0.6798477338109503,0.7672365450318365,0.9541375560379602,gwangju,best
+deepgbm+resnet_like,0.665680366173557,0.7714595133764687,0.9543727976423164,gwangju,best
+deepgbm+XGBoost,0.6574065703012586,0.7719143895834364,0.9543191075928837,gwangju,best
+deepgbm+LightGBM,0.6615432493953055,0.7728007686314036,0.9544770483485955,gwangju,best
+ft_transformer+resnet_like,0.5874777029243061,0.7627495358829616,0.9528777415974249,gwangju,best
+ft_transformer+XGBoost,0.5819133878684225,0.7549252510472145,0.9515179410587,gwangju,best
+ft_transformer+LightGBM,0.5805334443959108,0.7489469355258954,0.9505456293509993,gwangju,best
+resnet_like+XGBoost,0.5621088257073251,0.742686095459662,0.9492615719369842,gwangju,best
+resnet_like+LightGBM,0.5552708987061287,0.7370427725250878,0.9481963158420041,gwangju,best
+XGBoost+LightGBM,0.5288440675809284,0.7303546927519567,0.9467888046570956,gwangju,best
+deepgbm+ft_transformer+resnet_like,0.6605998838783642,0.7743701608802702,0.9553335870116691,gwangju,best
+deepgbm+ft_transformer+XGBoost,0.6495239084733958,0.7651993867637442,0.9536631567565769,gwangju,best
+deepgbm+ft_transformer+LightGBM,0.6556854096497969,0.7699591127202252,0.9550316682053713,gwangju,best
+deepgbm+resnet_like+XGBoost,0.646377396487214,0.7644362437422076,0.9526370070946761,gwangju,best
+deepgbm+resnet_like+LightGBM,0.6441717942778256,0.7623630196443353,0.9527517861450042,gwangju,best
+deepgbm+XGBoost+LightGBM,0.6104936580055047,0.7355161749128437,0.9470099225657277,gwangju,best
+ft_transformer+resnet_like+XGBoost,0.5926724219238438,0.7169506517948077,0.9453380367792,gwangju,best
+ft_transformer+resnet_like+LightGBM,0.5868504053718506,0.7120674745305696,0.9448050003742795,gwangju,best
+ft_transformer+XGBoost+LightGBM,0.5697749480416849,0.6978540261423789,0.9408133093794446,gwangju,best
+resnet_like+XGBoost+LightGBM,0.5568421888456356,0.6879639430352676,0.9383421289018639,gwangju,best
+deepgbm+ft_transformer+resnet_like+XGBoost,0.6464484806055241,0.7628617034260374,0.953282847185834,gwangju,best
+deepgbm+ft_transformer+resnet_like+LightGBM,0.6432761028976647,0.7603027833589299,0.9530166928662326,gwangju,best
+deepgbm+ft_transformer+XGBoost+LightGBM,0.6268590959555621,0.7470252829879999,0.9501275669336527,gwangju,best
+deepgbm+resnet_like+XGBoost+LightGBM,0.617990326312608,0.7412687534166494,0.9484927971987257,gwangju,best
+ft_transformer+resnet_like+XGBoost+LightGBM,0.5778772266115083,0.7046854387067157,0.9426776372150277,gwangju,best
+deepgbm+ft_transformer+resnet_like+XGBoost+LightGBM,0.6251709503819354,0.7454103687111592,0.9500901389824588,gwangju,best

Analysis_code/model_result/deepgbm_sampled_data_test.csv

@@ -0,0 +1,31 @@
+region,model,data_sample,CSI,MCC,Accuracy
+seoul,deepgbm,pure,0.6442509203188513,0.7660970126461858,0.9575257213197927
+busan,deepgbm,pure,0.6818642050717673,0.8049988566475562,0.9749078856534504
+incheon,deepgbm,pure,0.5671286321477068,0.6887072986816571,0.9090567324566875
+daegu,deepgbm,pure,0.5768403304029271,0.7158561147888857,0.9752356921259908
+daejeon,deepgbm,pure,0.6879666308365847,0.797628286801526,0.9573174763580109
+gwangju,deepgbm,pure,0.6169039999207967,0.7404995782864335,0.9513605060259002
+seoul,deepgbm,smote,0.6178386591753761,0.7437507767516012,0.9473890885046287
+busan,deepgbm,smote,0.5771081499427251,0.7328290895553372,0.9590890660478578
+incheon,deepgbm,smote,0.559840116480077,0.6829120896700739,0.9007683126647871
+daegu,deepgbm,smote,0.6425770288404345,0.7686582033950807,0.9787012085069575
+daejeon,deepgbm,smote,0.5569398315371836,0.6906810572279379,0.9299289492726501
+gwangju,deepgbm,smote,0.5526168770566259,0.6910361569617187,0.9341257662333341
+seoul,deepgbm,ctgan20000,0.7095252362606882,0.8058537207489471,0.9619258968152972
+busan,deepgbm,ctgan20000,0.5799669524449778,0.7334430424969706,0.9640024203408438
+incheon,deepgbm,ctgan20000,0.5289538894298621,0.6580659191964314,0.8935004283421081
+daegu,deepgbm,ctgan20000,0.44609227542135127,0.616365605274888,0.9682904159492977
+daejeon,deepgbm,ctgan20000,0.5716353901847903,0.7095774395951602,0.9385713751029269
+gwangju,deepgbm,ctgan20000,0.4571493095973662,0.620850435638757,0.9200951709625637
+seoul,deepgbm,ctgan10000,0.5616336216829804,0.700657840332931,0.9383062604486363
+busan,deepgbm,ctgan10000,0.5879352056252677,0.7351657312694786,0.9672767422711281
+incheon,deepgbm,ctgan10000,0.5290269463089048,0.6582877751557777,0.8928201753291581
+daegu,deepgbm,ctgan10000,0.5557972102197738,0.7032214449405951,0.9747075421480318
+daejeon,deepgbm,ctgan10000,0.5413249801262358,0.6761149432485588,0.9286541116683718
+gwangju,deepgbm,ctgan10000,0.41816087724732665,0.5735307327535618,0.9039835816054095
+seoul,deepgbm,ctgan7000,0.5813133937275518,0.7156872321691327,0.9403897497317663
+busan,deepgbm,ctgan7000,0.597435747844303,0.7454746315945583,0.9670328367891807
+incheon,deepgbm,ctgan7000,0.5123361043389418,0.646967905975133,0.8907633014779882
+daegu,deepgbm,ctgan7000,0.6146700981464019,0.7399778687018386,0.9766398640949504
+daejeon,deepgbm,ctgan7000,0.5581763081057168,0.6896808823936648,0.9308861691244354
+gwangju,deepgbm,ctgan7000,0.5798647223742114,0.7012366964380711,0.9421613394216134