{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "# Notebook 2: Probabilistic Reasoning & Naive Bayes Classifier\n",
        "\n",
        "In Chapter 3, we learned that real-world AI operates in an environment full of uncertainty, incomplete information, and noise. To make decisions, AI uses **Probabilistic Reasoning** and updates its beliefs using **Bayes' Rule**.\n",
        "\n",
        "In this notebook, we will write Python code to calculate real probabilities and build a text classifier."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Experiment 1: Updating Beliefs with Bayes' Rule (Predicting a Monsoon Storm)\n",
        "\n",
        "Suppose we observe dark clouds (`Clouds`) in our village sky. We want to know: what is the probability that a *Sunto* (`Storm`) is coming today?\n",
        "\n",
        "Bayes' Rule formula:\n",
        "`P(Storm | Clouds) = (P(Clouds | Storm) * P(Storm)) / P(Clouds)`"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {},
      "source": [
        "# Let us set up our known weather probabilities from past seasons\n",
        "P_storm = 0.20           # 20% of days have a storm during this season\n",
        "P_clouds_if_storm = 0.85 # If there is a storm, 85% of the time there are dark clouds first\n",
        "P_clouds = 0.35          # 35% of days have dark clouds overall\n",
        "\n",
        "# Calculate probability of a storm given we see clouds today\n",
        "P_storm_if_clouds = (P_clouds_if_storm * P_storm) / P_clouds\n",
        "\n",
        "print(f\"Probability of a Storm given dark clouds: {P_storm_if_clouds * 100:.1f}%\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {},
      "source": [
        "## Experiment 2: Naive Bayes Text Classification\n",
        "\n",
        "Let us build a simple classifier that automatically categorizes incoming community feedback or repair messages into either **'Water & Irrigation'** or **'Electricity & Power'**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {},
      "source": [
        "# Keyword dictionary for each category\n",
        "keywords = {\n",
        "    \"Water\": [\"pipe\", \"well\", \"naka\", \"kyari\", \"water\", \"leak\", \"canal\"],\n",
        "    \"Electricity\": [\"wire\", \"pole\", \"bijli\", \"current\", \"transformer\", \"light\"]\n",
        "}\n",
        "\n",
        "def classify_message(text):\n",
        "    words = text.lower().split()\n",
        "    score_water = sum(1 for w in words if w in keywords[\"Water\"])\n",
        "    score_elec = sum(1 for w in words if w in keywords[\"Electricity\"])\n",
        "    \n",
        "    if score_water > score_elec:\n",
        "        return \"Water & Irrigation Category\"\n",
        "    elif score_elec > score_water:\n",
        "        return \"Electricity & Power Category\"\n",
        "    else:\n",
        "        return \"General / Unknown Category\"\n",
        "\n",
        "# Test our classifier on village messages!\n",
        "test_msgs = [\n",
        "    \"the water pipe near the well has a leak\",\n",
        "    \"the bijli wire fallen from the pole\",\n",
        "    \"we need repair for the naka and canal\"\n",
        "]\n",
        "\n",
        "for msg in test_msgs:\n",
        "    print(f\"Message: '{msg}' -> Classified as: {classify_message(msg)}\")"
      ],
      "execution_count": null,
      "outputs": []
    }
  ],
  "metadata": {
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "name": "python",
      "version": "3"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 5
}