{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "kz8lLSv6mVQo"
   },
   "source": [
    "# **\ud83e\udd16 Data Analysis & Visualization**"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "jpASMyIQMaAq"
   },
   "source": [
    "## **1.** \ud83d\udce6 Install required packages"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "f48c8f8c",
    "outputId": "ca8b5ef6-ebc4-4fc6-9373-e1577253fd21"
   },
   "outputs": [],
   "source": [
    "!pip install pandas matplotlib seaborn numpy textblob faker transformers vaderSentiment\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "NZd99NpKkKyp"
   },
   "source": [
    "## **2.** \u2705\ufe0f Load & inspect input datasets"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "_JBLmm508Uq2"
   },
   "source": [
    "### *a. Initial setup*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "eBDXPQz18Xrs"
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import random"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "IL8lZbMm8m3k"
   },
   "source": [
    "### *b. \u270b\ud83c\udffb\ud83d\uded1\u26d4\ufe0f Create the df_reviews dataframe from the synthetic_book_reviews.csv file*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "fdgjghfO8uuq"
   },
   "outputs": [],
   "source": [
    "df_reviews = pd.read_csv(\"synthetic_book_reviews.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "N-Dl37J0HLhU"
   },
   "source": [
    "### *c. \u270b\ud83c\udffb\ud83d\uded1\u26d4\ufe0f Create the df_sales dataframe from the synthetic_sales_data.csv file*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "6XZs3P7fHgQe"
   },
   "outputs": [],
   "source": [
    "df_sales = pd.read_csv(\"synthetic_sales_data.csv\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "MUI3SkmyrGQo"
   },
   "source": [
    "### *d. \u270b\ud83c\udffb\ud83d\uded1\u26d4\ufe0f Visualize the first few lines of the two final datasets: df_reviews and df_sales*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "p8FdQFXErOqE",
    "outputId": "b6673cc0-bb1d-4221-dcef-e1f9dec0d6c8"
   },
   "outputs": [],
   "source": [
    "print(df_sales.head())\n",
    "print(df_reviews.head())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Y3oqGHsmrQzx"
   },
   "source": [
    "### *d. Run a quality check on the datasets*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "id": "VArQGPoKrfLm",
    "outputId": "16e94f18-4f60-44ba-919e-fc129f7f440e"
   },
   "outputs": [],
   "source": [
    "def quality_check(df, name=\"DataFrame\"):\n",
    "    print(f\"\\n\ud83d\udd0d Quality Check Report for: {name}\")\n",
    "    print(\"=\" * (25 + len(name)))\n",
    "\n",
    "    # Basic info\n",
    "    print(f\"\\n\ud83d\udccf Shape: {df.shape}\")\n",
    "    print(\"\\n\ud83d\udd20 Column Types:\")\n",
    "    print(df.dtypes)\n",
    "\n",
    "    # Missing values\n",
    "    print(\"\\n\u2753 Missing Values:\")\n",
    "    print(df.isnull().sum())\n",
    "\n",
    "    # Duplicates\n",
    "    duplicate_count = df.duplicated().sum()\n",
    "    print(f\"\\n\ud83d\udccb Duplicate Rows: {duplicate_count}\")\n",
    "\n",
    "    # Summary stats\n",
    "    print(\"\\n\ud83d\udcca Summary Statistics:\")\n",
    "    display(df.describe(include='all').transpose())\n",
    "\n",
    "    # Sample rows\n",
    "    print(\"\\n\ud83d\udc40 Sample Rows:\")\n",
    "    display(df.sample(5))\n",
    "\n",
    "# Run checks\n",
    "quality_check(df_reviews, \"df_reviews\")\n",
    "quality_check(df_sales, \"df_sales\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "TTxUKDYINPxV"
   },
   "source": [
    "## **3.** \ud83c\udfad Perform sentiment analysis using VADER"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "OqhYU8rDxQRT"
   },
   "source": [
    "### *a. Initial setup*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "DNk5w8mNxSZ6"
   },
   "outputs": [],
   "source": [
    "from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer\n",
    "\n",
    "# \ud83e\udd16 Initialize VADER analyzer\n",
    "analyzer = SentimentIntensityAnalyzer()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "P123TwSWxVAr"
   },
   "source": [
    "### *b. Create a function get_sentiment_label that will return the label negative, neutral, or positive based on the VADER analyzer's scoring of the text*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "89809e6f"
   },
   "outputs": [],
   "source": [
    "def get_sentiment_label(text):\n",
    "    score = analyzer.polarity_scores(text)[\"compound\"]\n",
    "    if score >= 0.05:\n",
    "        return \"positive\"\n",
    "    elif score <= -0.05:\n",
    "        return \"negative\"\n",
    "    else:\n",
    "        return \"neutral\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "DS9eCZ95yQn3"
   },
   "source": [
    "### *c. \u270b\ud83c\udffb\ud83d\uded1\u26d4\ufe0f Apply get_sentiment_label to df_reviews column named review_text to get sentiment_label column*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "SpXzFaDfyM7I"
   },
   "outputs": [],
   "source": [
    "df_reviews[\"sentiment_label\"] = df_reviews[\"review_text\"].apply(get_sentiment_label)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "5cnPCFFnyXN6"
   },
   "source": [
    "### *d. \u270b\ud83c\udffb\ud83d\uded1\u26d4\ufe0f View the first few lines of the resulting table df_reviews*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "ODGyfjBSyZEO",
    "outputId": "b18ef853-21d0-47d0-d9c6-3db2fcb641ac"
   },
   "outputs": [],
   "source": [
    "print(df_reviews.head())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Qy3Hqm-FojvT"
   },
   "source": [
    "## **4.** \ud83d\udcca Use the following data visualization code snippets"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "lcjGSw2bzqtZ"
   },
   "source": [
    "### *a. Initial setup*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "p5LV2o1rzsiC"
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import matplotlib.dates as mdates"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "tvaBtswpGS__",
    "outputId": "2800b619-e62e-4605-8a04-4a875859ef6b"
   },
   "outputs": [],
   "source": [
    "# ----------------------------\n",
    "# Outputs (for Hugging Face app)\n",
    "# ----------------------------\n",
    "# In the notebook: you still SEE interactive tables/plots inline.\n",
    "# For the Space dashboard: we also SAVE the same outputs as files.\n",
    "\n",
    "from pathlib import Path\n",
    "\n",
    "ART_DIR = Path(\"artifacts\")\n",
    "PY_FIG = ART_DIR / \"py\" / \"figures\"\n",
    "PY_TAB = ART_DIR / \"py\" / \"tables\"\n",
    "\n",
    "for p in [PY_FIG, PY_TAB]:\n",
    "    p.mkdir(parents=True, exist_ok=True)\n",
    "\n",
    "print(\"\u2705 Output folders:\")\n",
    "print(\" -\", PY_FIG.resolve())\n",
    "print(\" -\", PY_TAB.resolve())\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "b9T1rkBe0AJU"
   },
   "source": [
    "### *b. Sample of 5 books for each popularity level for visualizations*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "sLdFmGqXqo_t"
   },
   "outputs": [],
   "source": [
    "sampled_titles = []\n",
    "for pop_score in sorted(df_reviews[\"popularity_score\"].dropna().unique()):\n",
    "    all_titles = df_reviews[df_reviews[\"popularity_score\"] == pop_score][\"title\"].unique()\n",
    "    sampled = random.sample(list(all_titles), min(5, len(all_titles)))\n",
    "    sampled_titles.extend(sampled)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "xq7-C8m70mMH"
   },
   "source": [
    "### *c. Copy relevant sales, reviews, and book names*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "laDdMece0qrq"
   },
   "outputs": [],
   "source": [
    "sampled_sales = df_sales[df_sales[\"title\"].isin(sampled_titles)].copy()\n",
    "sampled_reviews = df_reviews[df_reviews[\"title\"].isin(sampled_titles)].copy()\n",
    "sampled_books = df_reviews[df_reviews[\"title\"].isin(sampled_titles)].copy()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "8YtfkG_A0wTy"
   },
   "source": [
    "### *d. Plot sales trends over time for the sampled books*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 0
    },
    "id": "1iTVzflW0Rkw",
    "outputId": "4193480a-0278-4a27-ad5d-3f7fcd700383"
   },
   "outputs": [],
   "source": [
    "# \ud83d\udd52 Ensure datetime format\n",
    "df_sales[\"month\"] = pd.to_datetime(df_sales[\"month\"])\n",
    "# \ud83c\udfa8 Color mapping\n",
    "popularity_colors = {\n",
    "    1: \"darkred\", 2: \"orangered\", 3: \"gold\", 4: \"mediumseagreen\", 5: \"royalblue\"\n",
    "}\n",
    "\n",
    "# \ud83d\udcc8 Plot 1: Sales trends\n",
    "plt.figure(figsize=(20, 8))\n",
    "for title in sampled_titles:\n",
    "    row = sampled_books[sampled_books[\"title\"] == title].iloc[0]\n",
    "    color = popularity_colors.get(row[\"popularity_score\"], \"gray\")\n",
    "    subset = sampled_sales[sampled_sales[\"title\"] == title]\n",
    "    plt.plot(subset[\"month\"], subset[\"units_sold\"], label=f\"{title} (Pop. {row['popularity_score']})\", color=color)\n",
    "\n",
    "plt.title(\"\ud83d\udcc8 Sales Trends Over Time (5 per Popularity Level)\")\n",
    "plt.xlabel(\"Month\")\n",
    "plt.ylabel(\"Units Sold\")\n",
    "plt.xticks(rotation=45)\n",
    "plt.legend(loc='center left', bbox_to_anchor=(1, 0.5), fontsize='small')\n",
    "plt.grid(True)\n",
    "plt.tight_layout()\n",
    "plt.gca().xaxis.set_major_formatter(mdates.DateFormatter('%b %Y'))\n",
    "plt.savefig(PY_FIG / 'sales_trends_sampled_titles.png', dpi=150)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "lDpMkjDP1K6j"
   },
   "source": [
    "### *e. Plot sentiment_label distribution per book*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 0
    },
    "id": "dn1Jgd5R1KLu",
    "outputId": "78319480-be56-40d7-8935-c5745d9e37ed"
   },
   "outputs": [],
   "source": [
    "# \ud83c\udfa8 Give a new name to each book that includes the rating together with the title\n",
    "sampled_reviews[\"grouped_title\"] = sampled_reviews[\"rating\"].astype(str) + \"\u2605 | \" + sampled_reviews[\"title\"]\n",
    "\n",
    "# \ud83d\udcca Aggregate sentiment counts\n",
    "sentiment_counts = (\n",
    "    sampled_reviews.groupby([\"grouped_title\", \"sentiment_label\"])\n",
    "    .size()\n",
    "    .unstack(fill_value=0)[[\"negative\", \"neutral\", \"positive\"]]  # consistent order\n",
    ")\n",
    "\n",
    "# \ud83d\udcbe Save table for HF dashboard\n",
    "sentiment_counts.reset_index().to_csv(PY_TAB / 'sentiment_counts_sampled.csv', index=False)\n",
    "\n",
    "\n",
    "# \u2705 Plot stacked horizontal bars\n",
    "fig, ax = plt.subplots(figsize=(12, 14))\n",
    "sentiment_counts.plot.barh(\n",
    "    stacked=True,\n",
    "    ax=ax,\n",
    "    color={\"negative\": \"royalblue\", \"neutral\": \"lightgray\", \"positive\": \"crimson\"}\n",
    ")\n",
    "\n",
    "plt.title(\"\ud83d\udcac Sentiment Distribution in Reviews (5 Books per Popularity Level)\", fontsize=14)\n",
    "plt.xlabel(\"Number of Reviews\")\n",
    "plt.ylabel(\"Book Title (Grouped by Popularity Score)\")\n",
    "plt.legend(title=\"Sentiment\", loc=\"lower right\")\n",
    "plt.grid(axis=\"x\", linestyle=\"--\", alpha=0.6)\n",
    "plt.tight_layout()\n",
    "plt.savefig(PY_FIG / 'sentiment_distribution_sampled_titles.png', dpi=150)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "rmgylC1ENCHy"
   },
   "source": [
    "## **5.** \ud83d\udd2e Forecast book sales with the following  ARIMA code"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "jFV4JE1R3FKH"
   },
   "source": [
    "### *a. Initial setup*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "Mh8Alha03H22"
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.dates as mdates\n",
    "import statsmodels.api as sm\n",
    "from itertools import product\n",
    "import matplotlib.cm as cm\n",
    "import warnings"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "gHucD8OW3U0w"
   },
   "source": [
    "### *b. Define function find_best_arima to try different ARIMA parameter values and return the best combination for each book's price forecast*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "477fa43f"
   },
   "outputs": [],
   "source": [
    "def find_best_arima(series, p_range=(0, 5), d_range=(0, 2), q_range=(0, 1)):\n",
    "    best_aic = float(\"inf\")\n",
    "    best_order = None\n",
    "    best_model = None\n",
    "\n",
    "    for p, d, q in product(range(p_range[0], p_range[1] + 1),\n",
    "                           range(d_range[0], d_range[1] + 1),\n",
    "                           range(q_range[0], q_range[1] + 1)):\n",
    "        try:\n",
    "            model = sm.tsa.ARIMA(series, order=(p, d, q))\n",
    "            results = model.fit()\n",
    "            if results.aic < best_aic:\n",
    "                best_aic = results.aic\n",
    "                best_order = (p, d, q)\n",
    "                best_model = results\n",
    "        except:\n",
    "            continue\n",
    "\n",
    "    return best_order, best_model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Rq5t1Hey3jkD"
   },
   "source": [
    "### *c. Plot the figure*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 0
    },
    "id": "DmxGdvLE3dHQ",
    "outputId": "9bb19d6b-2f68-4066-cbec-dd7eb80a9111"
   },
   "outputs": [],
   "source": [
    "# \ud83c\udfa8 Generate 25 highly distinct colors using HUSL (hue-saturation-lightness)\n",
    "colors = sns.color_palette(\"tab10\", len(sampled_titles))\n",
    "\n",
    "plt.figure(figsize=(16, 10))\n",
    "\n",
    "for i, title in enumerate(sampled_titles):\n",
    "    book_sales = sampled_sales[sampled_sales[\"title\"] == title].copy()\n",
    "    book_sales[\"month\"] = pd.to_datetime(book_sales[\"month\"])\n",
    "    book_sales = book_sales.sort_values(\"month\").set_index(\"month\")\n",
    "\n",
    "    with warnings.catch_warnings():\n",
    "        warnings.simplefilter(\"ignore\")\n",
    "        best_order, best_model = find_best_arima(book_sales[\"units_sold\"])\n",
    "        if best_model is not None:\n",
    "            forecast = best_model.get_forecast(steps=6)\n",
    "            forecast_index = pd.date_range(start=book_sales.index[-1] + pd.DateOffset(months=1), periods=6, freq='MS')\n",
    "\n",
    "            # \ud83d\udfe6 Plot observed sales (solid line)\n",
    "            plt.plot(book_sales.index, book_sales[\"units_sold\"], color=colors[i], label=title, linewidth=2)\n",
    "\n",
    "            # \ud83d\udfe0 Plot forecast (dotted line, same color)\n",
    "            plt.plot(forecast_index, forecast.predicted_mean, linestyle=\"--\", color=colors[i], linewidth=2)\n",
    "\n",
    "# \ud83d\udcc8 Final formatting\n",
    "plt.title(\"\ud83d\udcc8 ARIMA Forecasts for Sampled Books (1 per Popularity Level)\", fontsize=14)\n",
    "plt.xlabel(\"Month\")\n",
    "plt.ylabel(\"Units Sold\")\n",
    "plt.xticks(rotation=45)\n",
    "plt.gca().xaxis.set_major_formatter(mdates.DateFormatter('%b %Y'))\n",
    "plt.grid(True)\n",
    "plt.legend(loc=\"center left\", bbox_to_anchor=(1, 0.5), fontsize=\"small\")\n",
    "plt.tight_layout()\n",
    "plt.savefig(PY_FIG / 'arima_forecasts_sampled_titles.png', dpi=150)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "SKBcx3fyCFly"
   },
   "source": [
    "## **6.** \ud83c\udff7\ufe0f Decide on price changes with a rule-based approach based on sentiment and future revenue"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "nY-vV2JJDZqu"
   },
   "source": [
    "### *a. Calculate average sales per book*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "nbDT_RHaDD2R"
   },
   "outputs": [],
   "source": [
    "avg_sales = df_sales.groupby(\"title\")[\"units_sold\"].mean().reset_index()\n",
    "avg_sales.columns = [\"title\", \"avg_units_sold\"]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "94wi-RvkDf2z"
   },
   "source": [
    "### *b. Calculate sentiment distribution per book*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "fWjQ9IOXDk-M"
   },
   "outputs": [],
   "source": [
    "sentiment_counts = df_reviews.groupby([\"title\", \"sentiment_label\"]).size().unstack(fill_value=0)\n",
    "sentiment_counts[\"total\"] = sentiment_counts.sum(axis=1)\n",
    "sentiment_counts[\"positive_ratio\"] = sentiment_counts.get(\"positive\") / sentiment_counts[\"total\"]\n",
    "sentiment_counts[\"negative_ratio\"] = sentiment_counts.get(\"negative\") / sentiment_counts[\"total\"]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Vm10ym_iDtEW"
   },
   "source": [
    "### *c. Merge the calculated sales and sentiment characteristics*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "T-zlh6rBDpxg"
   },
   "outputs": [],
   "source": [
    "df_decision = avg_sales.merge(sentiment_counts, on=\"title\", how=\"left\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "1WIWDojyD7fK"
   },
   "source": [
    "### *d. \u270b\ud83c\udffb\ud83d\uded1\u26d4\ufe0f Create the pricing_decision function as a basic rule-based pricing decider based on sentiment and revenue*\n",
    "\n",
    "\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "b5qJCb46Dxfb"
   },
   "source": [
    "*   If there are 120 or more average units sold and 0.6 or higher positive ratio, the decision should be to increase price.\n",
    "*   If there are 60 or less average units sold and 0.4 or higher negative ratio, the decision should be to decrease price.\n",
    "*   Otherwise, the price should be kept the same."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "XBzozedwD6yx"
   },
   "outputs": [],
   "source": [
    "def pricing_decision(row):\n",
    "    if row[\"avg_units_sold\"] >= 120 and row[\"positive_ratio\"] >= 0.6:\n",
    "        return \"Increase Price\"\n",
    "    elif row[\"avg_units_sold\"] <= 60 and row[\"negative_ratio\"] >= 0.4:\n",
    "        return \"Decrease Price\"\n",
    "    else:\n",
    "        return \"Keep Price\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "xmLEdF14EPAA"
   },
   "source": [
    "### *e. \u270b\ud83c\udffb\ud83d\uded1\u26d4\ufe0f Run the pricing_decision function and check out the first few decisions*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "TZ0ZhgHrEQJB",
    "outputId": "7a7dd343-32b5-47f4-f8b3-f9581fe52942"
   },
   "outputs": [],
   "source": [
    "df_decision[\"pricing_action\"] = df_decision.apply(pricing_decision, axis=1)\n",
    "\n",
    "print(df_decision.head())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "WTkP2_-EApev"
   },
   "source": [
    "\n",
    "## **7.** \ud83d\udcbe Save Python outputs for the Hugging Face dashboard"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "3EIjfnokGpJv"
   },
   "source": [
    "\n",
    "This section exports **HF-ready artifacts** into a consistent folder structure:\n",
    "\n",
    "- `(root folder)py/figures/` (Python-generated visuals)\n",
    "- `(root folder)py/tables/` (tables/metrics)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "ZJJ4PMgIApev",
    "outputId": "159cb98e-456f-4f6f-a73c-06ed396f1fb8"
   },
   "outputs": [],
   "source": [
    "\n",
    "import json\n",
    "\n",
    "# -------------------------\n",
    "# 1) Dashboard table (monthly) \u2014 reuse if already built\n",
    "# -------------------------\n",
    "if \"df_monthly\" in globals() and df_monthly is not None:\n",
    "    df_dashboard = df_monthly.copy()\n",
    "else:\n",
    "    # fallback: monthly units sold only\n",
    "    df_dashboard = (\n",
    "        df_sales.groupby(\"month\", as_index=False)\n",
    "        .agg(total_units_sold=(\"units_sold\", \"sum\"))\n",
    "        .sort_values(\"month\")\n",
    "    )\n",
    "\n",
    "# Save the single overview dashboard table\n",
    "df_dashboard.to_csv(PY_TAB / \"df_dashboard.csv\", index=False)\n",
    "\n",
    "# -------------------------\n",
    "# 2) KPI summary (small json) \u2014 computed from raw df_sales + df_dashboard\n",
    "# -------------------------\n",
    "kpis = {\n",
    "    \"n_titles\": int(df_sales[\"title\"].nunique()),\n",
    "    \"n_months\": int(df_dashboard[\"month\"].nunique()),\n",
    "    \"total_units_sold\": float(df_sales[\"units_sold\"].sum()),\n",
    "}\n",
    "\n",
    "# Only include revenue KPIs if df_dashboard contains it (since you said monthly revenue already exists)\n",
    "if \"total_revenue\" in df_dashboard.columns and df_dashboard[\"total_revenue\"].notna().any():\n",
    "    kpis[\"total_revenue\"] = float(df_dashboard[\"total_revenue\"].sum())\n",
    "\n",
    "with open(PY_FIG / \"kpis.json\", \"w\", encoding=\"utf-8\") as f:\n",
    "    json.dump(kpis, f, indent=2)\n",
    "\n",
    "# -------------------------\n",
    "# 3) Python tables (title-level quick inspection)\n",
    "# -------------------------\n",
    "df_by_title_units = (\n",
    "    df_sales.groupby(\"title\", as_index=False)\n",
    "    .agg(total_units_sold=(\"units_sold\", \"sum\"))\n",
    "    .sort_values(\"total_units_sold\", ascending=False)\n",
    ")\n",
    "df_by_title_units.head(10).to_csv(PY_TAB / \"top_titles_by_units_sold.csv\", index=False)\n",
    "\n",
    "# Optional: title-level revenue table ONLY if df_sales already has per-row revenue\n",
    "if \"revenue\" in df_sales.columns and df_sales[\"revenue\"].notna().any():\n",
    "    df_by_title_rev = (\n",
    "        df_sales.groupby(\"title\", as_index=False)\n",
    "        .agg(total_revenue=(\"revenue\", \"sum\"))\n",
    "        .sort_values(\"total_revenue\", ascending=False)\n",
    "    )\n",
    "    df_by_title_rev.head(10).to_csv(PY_TAB / \"top_titles_by_revenue.csv\", index=False)\n",
    "\n",
    "print(\"\u2705 Exports written to artifacts/:\")\n",
    "print(\" - common/: df_dashboard.csv, kpis.json\")\n",
    "print(\" - py/tables/: top_titles_by_units_sold.csv (+ optional top_titles_by_revenue.csv)\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "0b4e76d3"
   },
   "source": [
    "\u2705 **Extra outputs for the R notebook**: `(root folder)common/r_input_title_level.csv` and `(root folder)common/r_input_monthly_revenue.csv` (these are the only two files the R portion needs)."
   ]
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [
    "jpASMyIQMaAq",
    "NZd99NpKkKyp",
    "TTxUKDYINPxV",
    "Qy3Hqm-FojvT",
    "rmgylC1ENCHy",
    "SKBcx3fyCFly",
    "WTkP2_-EApev"
   ],
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "name": "python3"
  },
  "language_info": {
   "name": "python"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}