Delete ranalysis.ipynb

ranalysis.ipynb

@@ -1,463 +0,0 @@
-{
-  "cells": [
-    {
-      "cell_type": "markdown",
-      "id": "75fd9cc6",
-      "metadata": {
-        "id": "75fd9cc6"
-      },
-      "source": [
-        "# **🤖 Benchmarking & Modeling**"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "id": "fb807724",
-      "metadata": {
-        "id": "fb807724"
-      },
-      "source": [
-        "## **1.** 📦 Setup"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "id": "d40cd131",
-      "metadata": {
-        "id": "d40cd131"
-      },
-      "outputs": [],
-      "source": [
-        "\n",
-        "# Uncomment the next line once:\n",
-        "install.packages(c(\"readr\",\"dplyr\",\"stringr\",\"tidyr\",\"lubridate\",\"ggplot2\",\"forecast\",\"broom\",\"jsonlite\"), repos=\"https://cloud.r-project.org\")\n",
-        "\n",
-        "suppressPackageStartupMessages({\n",
-        "  library(readr)\n",
-        "  library(dplyr)\n",
-        "  library(stringr)\n",
-        "  library(tidyr)\n",
-        "  library(lubridate)\n",
-        "  library(ggplot2)\n",
-        "  library(forecast)\n",
-        "  library(broom)\n",
-        "  library(jsonlite)\n",
-        "})"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "id": "f01d02e7",
-      "metadata": {
-        "id": "f01d02e7"
-      },
-      "source": [
-        "## **2.** ✅️ Load & inspect inputs"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "id": "29e8f6ce",
-      "metadata": {
-        "colab": {
-          "base_uri": "https://localhost:8080/"
-        },
-        "id": "29e8f6ce",
-        "outputId": "5a1bda1c-c58d-43d0-c85e-db5041c8bc49"
-      },
-      "outputs": [
-        {
-          "output_type": "stream",
-          "name": "stdout",
-          "text": [
-            "Loaded: 1000 rows (title-level), 18 rows (monthly)\n"
-          ]
-        }
-      ],
-      "source": [
-        "\n",
-        "must_exist <- function(path, label) {\n",
-        "  if (!file.exists(path)) stop(paste0(\"Missing \", label, \": \", path))\n",
-        "}\n",
-        "\n",
-        "TITLE_PATH <- \"synthetic_title_level_features.csv\"\n",
-        "MONTH_PATH <- \"synthetic_monthly_revenue_series.csv\"\n",
-        "\n",
-        "must_exist(TITLE_PATH, \"TITLE_PATH\")\n",
-        "must_exist(MONTH_PATH, \"MONTH_PATH\")\n",
-        "\n",
-        "df_title <- read_csv(TITLE_PATH, show_col_types = FALSE)\n",
-        "df_month <- read_csv(MONTH_PATH, show_col_types = FALSE)\n",
-        "\n",
-        "cat(\"Loaded:\", nrow(df_title), \"rows (title-level),\", nrow(df_month), \"rows (monthly)\n",
-        "\")"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "id": "9fd04262",
-      "metadata": {
-        "colab": {
-          "base_uri": "https://localhost:8080/"
-        },
-        "id": "9fd04262",
-        "outputId": "5f031538-96be-4758-904d-9201ec3c3ea7"
-      },
-      "outputs": [
-        {
-          "output_type": "stream",
-          "name": "stdout",
-          "text": [
-            "\u001b[90m# A tibble: 1 × 6\u001b[39m\n",
-            "      n na_avg_revenue na_price na_rating na_share_pos na_share_neg\n",
-            "  \u001b[3m\u001b[90m<int>\u001b[39m\u001b[23m          \u001b[3m\u001b[90m<int>\u001b[39m\u001b[23m    \u001b[3m\u001b[90m<int>\u001b[39m\u001b[23m     \u001b[3m\u001b[90m<int>\u001b[39m\u001b[23m        \u001b[3m\u001b[90m<int>\u001b[39m\u001b[23m        \u001b[3m\u001b[90m<int>\u001b[39m\u001b[23m\n",
-            "\u001b[90m1\u001b[39m  \u001b[4m1\u001b[24m000              0        0      \u001b[4m1\u001b[24m000            0            0\n",
-            "Monthly rows after parsing: 18 \n"
-          ]
-        }
-      ],
-      "source": [
-        "\n",
-        "# ---------- helpers ----------\n",
-        "safe_num <- function(x) {\n",
-        "  # strips anything that is not digit or dot\n",
-        "  suppressWarnings(as.numeric(str_replace_all(as.character(x), \"[^0-9.]\", \"\")))\n",
-        "}\n",
-        "\n",
-        "parse_rating <- function(x) {\n",
-        "  # Accept: 4, \"4\", \"4.0\", \"4/5\", \"4 out of 5\", \"⭐⭐⭐⭐\", etc.\n",
-        "  x <- as.character(x)\n",
-        "  x <- str_replace_all(x, \"⭐\", \"\")\n",
-        "  x <- str_to_lower(x)\n",
-        "  x <- str_replace_all(x, \"stars?\", \"\")\n",
-        "  x <- str_replace_all(x, \"out of\", \"/\")\n",
-        "  x <- str_replace_all(x, \"\\\\s+\", \"\")\n",
-        "  x <- str_replace_all(x, \"[^0-9./]\", \"\")\n",
-        "  suppressWarnings(as.numeric(str_extract(x, \"^[0-9.]+\")))\n",
-        "}\n",
-        "\n",
-        "parse_month <- function(x) {\n",
-        "  x <- as.character(x)\n",
-        "  # try YYYY-MM-DD, then YYYY-MM\n",
-        "  out <- suppressWarnings(ymd(x))\n",
-        "  if (mean(is.na(out)) > 0.5) out <- suppressWarnings(ymd(paste0(x, \"-01\")))\n",
-        "  na_idx <- which(is.na(out))\n",
-        "  if (length(na_idx) > 0) out[na_idx] <- suppressWarnings(ymd(paste0(x[na_idx], \"-01\")))\n",
-        "  out\n",
-        "}\n",
-        "\n",
-        "# ---------- normalize keys ----------\n",
-        "df_title <- df_title %>% mutate(title = str_squish(as.character(title)))\n",
-        "df_month <- df_month %>% mutate(month = as.character(month))\n",
-        "\n",
-        "# ---------- parse numeric columns defensively ----------\n",
-        "need_cols_title <- c(\"title\",\"avg_revenue\",\"total_revenue\",\"price\",\"rating\",\"share_positive\",\"share_negative\",\"share_neutral\")\n",
-        "missing_title <- setdiff(need_cols_title, names(df_title))\n",
-        "if (length(missing_title) > 0) stop(paste0(\"df_title missing columns: \", paste(missing_title, collapse=\", \")))\n",
-        "\n",
-        "df_title <- df_title %>%\n",
-        "  mutate(\n",
-        "    avg_revenue    = safe_num(avg_revenue),\n",
-        "    total_revenue  = safe_num(total_revenue),\n",
-        "    price          = safe_num(price),\n",
-        "    rating         = parse_rating(rating),\n",
-        "    share_positive = safe_num(share_positive),\n",
-        "    share_negative = safe_num(share_negative),\n",
-        "    share_neutral  = safe_num(share_neutral)\n",
-        "  )\n",
-        "\n",
-        "# basic sanity stats\n",
-        "hyg <- df_title %>%\n",
-        "  summarise(\n",
-        "    n = n(),\n",
-        "    na_avg_revenue = sum(is.na(avg_revenue)),\n",
-        "    na_price = sum(is.na(price)),\n",
-        "    na_rating = sum(is.na(rating)),\n",
-        "    na_share_pos = sum(is.na(share_positive)),\n",
-        "    na_share_neg = sum(is.na(share_negative))\n",
-        "  )\n",
-        "\n",
-        "print(hyg)\n",
-        "\n",
-        "# monthly parsing\n",
-        "need_cols_month <- c(\"month\",\"total_revenue\")\n",
-        "missing_month <- setdiff(need_cols_month, names(df_month))\n",
-        "if (length(missing_month) > 0) stop(paste0(\"df_month missing columns: \", paste(missing_month, collapse=\", \")))\n",
-        "\n",
-        "df_month2 <- df_month %>%\n",
-        "  mutate(\n",
-        "    month = parse_month(month),\n",
-        "    total_revenue = safe_num(total_revenue)\n",
-        "  ) %>%\n",
-        "  filter(!is.na(month)) %>%\n",
-        "  arrange(month)\n",
-        "\n",
-        "cat(\"Monthly rows after parsing:\", nrow(df_month2), \"\\n\")"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "id": "b8971bc4",
-      "metadata": {
-        "id": "b8971bc4"
-      },
-      "source": [
-        "## **3.** 💾 Folder for R outputs for Hugging Face"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "id": "dfaa06b1",
-      "metadata": {
-        "colab": {
-          "base_uri": "https://localhost:8080/"
-        },
-        "id": "dfaa06b1",
-        "outputId": "73f6437a-39f4-4968-f88a-99f10a3fd8ae"
-      },
-      "outputs": [
-        {
-          "output_type": "stream",
-          "name": "stdout",
-          "text": [
-            "R outputs will be written to: /content/artifacts/r \n"
-          ]
-        }
-      ],
-      "source": [
-        "\n",
-        "ART_DIR  <- \"artifacts\"\n",
-        "R_FIG_DIR <- file.path(ART_DIR, \"r\", \"figures\")\n",
-        "R_TAB_DIR <- file.path(ART_DIR, \"r\", \"tables\")\n",
-        "\n",
-        "dir.create(R_FIG_DIR, recursive = TRUE, showWarnings = FALSE)\n",
-        "dir.create(R_TAB_DIR, recursive = TRUE, showWarnings = FALSE)\n",
-        "\n",
-        "cat(\"R outputs will be written to:\", normalizePath(file.path(ART_DIR, \"r\"), winslash = \"/\"), \"\n",
-        "\")"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "id": "f880c72d",
-      "metadata": {
-        "id": "f880c72d"
-      },
-      "source": [
-        "## **4.** 🔮 Forecast book sales benchmarking with `accuracy()`"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "source": [
-        "We benchmark **three** models on a holdout window (last *h* months):\n",
-        "- ARIMA + Fourier (seasonality upgrade)\n",
-        "- ETS\n",
-        "- Naive baseline\n",
-        "\n",
-        "Then we export:\n",
-        "- `accuracy_table.csv`\n",
-        "- `forecast_compare.png`\n",
-        "- `rmse_comparison.png`"
-      ],
-      "metadata": {
-        "id": "R0JZlzKegmzW"
-      },
-      "id": "R0JZlzKegmzW"
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "id": "62e87992",
-      "metadata": {
-        "colab": {
-          "base_uri": "https://localhost:8080/"
-        },
-        "id": "62e87992",
-        "outputId": "73b36487-a25d-4bb9-cf80-8d5a654a2f0d"
-      },
-      "outputs": [
-        {
-          "output_type": "stream",
-          "name": "stdout",
-          "text": [
-            "✅ Saved: artifacts/r/tables/accuracy_table.csv\n",
-            "✅ Saved: artifacts/r/figures/rmse_comparison.png\n"
-          ]
-        },
-        {
-          "output_type": "display_data",
-          "data": {
-            "text/html": [
-              "<strong>agg_record_872216040:</strong> 2"
-            ],
-            "text/markdown": "**agg_record_872216040:** 2",
-            "text/latex": "\\textbf{agg\\textbackslash{}\\_record\\textbackslash{}\\_872216040:} 2",
-            "text/plain": [
-              "agg_record_872216040 \n",
-              "                   2 "
-            ]
-          },
-          "metadata": {}
-        },
-        {
-          "output_type": "stream",
-          "name": "stdout",
-          "text": [
-            "✅ Saved: artifacts/r/figures/forecast_compare.png\n"
-          ]
-        }
-      ],
-      "source": [
-        "\n",
-        "# Build monthly ts\n",
-        "start_year <- year(min(df_month2$month, na.rm = TRUE))\n",
-        "start_mon  <- month(min(df_month2$month, na.rm = TRUE))\n",
-        "\n",
-        "y <- ts(df_month2$total_revenue, frequency = 12, start = c(start_year, start_mon))\n",
-        "\n",
-        "# holdout size: min(6, 20% of series), at least 1\n",
-        "h_test <- min(6, max(1, floor(length(y) / 5)))\n",
-        "train_ts <- head(y, length(y) - h_test)\n",
-        "test_ts  <- tail(y, h_test)\n",
-        "\n",
-        "# Model A: ARIMA + Fourier\n",
-        "K <- 2\n",
-        "xreg_train <- fourier(train_ts, K = K)\n",
-        "fit_arima <- auto.arima(train_ts, xreg = xreg_train)\n",
-        "xreg_future <- fourier(train_ts, K = K, h = h_test)\n",
-        "fc_arima <- forecast(fit_arima, xreg = xreg_future, h = h_test)\n",
-        "\n",
-        "# Model B: ETS\n",
-        "fit_ets <- ets(train_ts)\n",
-        "fc_ets  <- forecast(fit_ets, h = h_test)\n",
-        "\n",
-        "# Model C: Naive baseline\n",
-        "fc_naive <- naive(train_ts, h = h_test)\n",
-        "\n",
-        "# accuracy() tables\n",
-        "acc_arima <- as.data.frame(accuracy(fc_arima, test_ts))\n",
-        "acc_ets   <- as.data.frame(accuracy(fc_ets, test_ts))\n",
-        "acc_naive <- as.data.frame(accuracy(fc_naive, test_ts))\n",
-        "\n",
-        "accuracy_tbl <- bind_rows(\n",
-        "  acc_arima %>% mutate(model = \"ARIMA+Fourier\"),\n",
-        "  acc_ets   %>% mutate(model = \"ETS\"),\n",
-        "  acc_naive %>% mutate(model = \"Naive\")\n",
-        ") %>% relocate(model)\n",
-        "\n",
-        "write_csv(accuracy_tbl, file.path(R_TAB_DIR, \"accuracy_table.csv\"))\n",
-        "cat(\"✅ Saved: artifacts/r/tables/accuracy_table.csv\\n\")\n",
-        "\n",
-        "# RMSE bar chart\n",
-        "p_rmse <- ggplot(accuracy_tbl, aes(x = reorder(model, RMSE), y = RMSE)) +\n",
-        "  geom_col() +\n",
-        "  coord_flip() +\n",
-        "  labs(title = \"Forecast model comparison (RMSE on holdout)\", x = \"\", y = \"RMSE\") +\n",
-        "  theme_minimal()\n",
-        "\n",
-        "ggsave(file.path(R_FIG_DIR, \"rmse_comparison.png\"), p_rmse, width = 8, height = 4, dpi = 160)\n",
-        "cat(\"✅ Saved: artifacts/r/figures/rmse_comparison.png\\n\")\n",
-        "\n",
-        "# Side-by-side forecast plots (simple, no extra deps)\n",
-        "png(file.path(R_FIG_DIR, \"forecast_compare.png\"), width = 1200, height = 500)\n",
-        "par(mfrow = c(1, 3))\n",
-        "plot(fc_arima, main = \"ARIMA + Fourier\", xlab = \"Time\", ylab = \"Total revenue\"); lines(test_ts, col = \"black\")\n",
-        "plot(fc_ets, main = \"ETS\", xlab = \"Time\", ylab = \"Total revenue\"); lines(test_ts, col = \"black\")\n",
-        "plot(fc_naive, main = \"Naive\", xlab = \"Time\", ylab = \"Total revenue\"); lines(test_ts, col = \"black\")\n",
-        "dev.off()\n",
-        "cat(\"✅ Saved: artifacts/r/figures/forecast_compare.png\\n\")"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "id": "30bc017b",
-      "metadata": {
-        "id": "30bc017b"
-      },
-      "source": [
-        "## **5.** 💾 Some R metadata for Hugging Face"
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "id": "645cb12b",
-      "metadata": {
-        "colab": {
-          "base_uri": "https://localhost:8080/"
-        },
-        "id": "645cb12b",
-        "outputId": "c00c26da-7d27-4c78-a296-aa33807495d4"
-      },
-      "outputs": [
-        {
-          "output_type": "stream",
-          "name": "stdout",
-          "text": [
-            "✅ Saved: artifacts/r/tables/r_meta.json\n",
-            "DONE. R artifacts written to: artifacts/r \n"
-          ]
-        }
-      ],
-      "source": [
-        "# =========================================================\n",
-        "# Metadata export (aligned with current notebook objects)\n",
-        "# =========================================================\n",
-        "\n",
-        "meta <- list(\n",
-        "\n",
-        "  # ---------------------------\n",
-        "  # Dataset footprint\n",
-        "  # ---------------------------\n",
-        "  n_titles = nrow(df_title),\n",
-        "  n_months = nrow(df_month2),\n",
-        "\n",
-        "  # ---------------------------\n",
-        "  # Forecasting info\n",
-        "  # (only if these objects exist in your forecasting section)\n",
-        "  # ---------------------------\n",
-        "  forecasting = list(\n",
-        "    holdout_h = h_test,\n",
-        "    arima_order = forecast::arimaorder(fit_arima),\n",
-        "    ets_method = fit_ets$method\n",
-        "  )\n",
-        ")\n",
-        "\n",
-        "jsonlite::write_json(\n",
-        "  meta,\n",
-        "  path = file.path(R_TAB_DIR, \"r_meta.json\"),\n",
-        "  pretty = TRUE,\n",
-        "  auto_unbox = TRUE\n",
-        ")\n",
-        "\n",
-        "cat(\"✅ Saved: artifacts/r/tables/r_meta.json\\n\")\n",
-        "cat(\"DONE. R artifacts written to:\", file.path(ART_DIR, \"r\"), \"\\n\")\n"
-      ]
-    }
-  ],
-  "metadata": {
-    "colab": {
-      "provenance": [],
-      "collapsed_sections": [
-        "f01d02e7",
-        "b8971bc4",
-        "f880c72d",
-        "30bc017b"
-      ]
-    },
-    "kernelspec": {
-      "name": "ir",
-      "display_name": "R"
-    },
-    "language_info": {
-      "name": "R"
-    }
-  },
-  "nbformat": 4,
-  "nbformat_minor": 5
-}