{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "457956e0",
   "metadata": {},
   "source": [
    "# Lesson 28: TensorFlow/Keras classification activity\n",
    "\n",
    "## Activity Goal\n",
    "\n",
    "In this activity, you will build and train a neural network using TensorFlow/Keras to predict room occupancy based on environmental sensor data. The activity will guide you through data preparation, creating a logistic regression baseline for comparison, and then building a neural network using Keras' Sequential API to see how deep learning performs on this binary classification task.\n",
    "\n",
    "## Notebook set up\n",
    "\n",
    "### Vocareum compatibility\n",
    "\n",
    "Check the notebook path to determine if we are running in Vocareum, if so, install TensorFlow 2.17 and Keras 3.2.0 via pip."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "88591257",
   "metadata": {},
   "outputs": [],
   "source": [
    "from pathlib import Path\n",
    "\n",
    "# Get the parts of the current directory path as a tuple\n",
    "path_parts = Path.cwd().parts\n",
    "\n",
    "# Check if 'voc' is the first directory level\n",
    "is_voc_root = len(path_parts) > 1 and path_parts[1] == \"voc\"\n",
    "\n",
    "# If we are in a Vocareum instance, upgrade TensorFlow and Keras\n",
    "if is_voc_root:\n",
    "    print(\"Running in Vocareum environment\")\n",
    "\n",
    "    get_ipython().run_line_magic(\n",
    "        \"pip\", \"install -qq --no-warn-script-location tensorflow==2.17.0 keras==3.2.0\"\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c20db720",
   "metadata": {},
   "source": [
    "### Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1ca8575f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Suppress noisy TensorFlow logs before importing\n",
    "import os\n",
    "os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'\n",
    "\n",
    "# Third party imports\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.metrics import accuracy_score, classification_report, ConfusionMatrixDisplay\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import layers"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e749a9df",
   "metadata": {},
   "source": [
    "### GPU configuration (if present)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4bcea4ad",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Configure GPU settings\n",
    "gpus = tf.config.list_physical_devices('GPU')\n",
    "\n",
    "if gpus:\n",
    "    try:\n",
    "        # Use only GPU 0\n",
    "        tf.config.set_visible_devices(gpus[0], 'GPU')\n",
    "\n",
    "        # Enable memory growth\n",
    "        tf.config.experimental.set_memory_growth(gpus[0], True)\n",
    "\n",
    "        # Get GPU\n",
    "        details = tf.config.experimental.get_device_details(gpus[0])\n",
    "        cc = details.get('compute_capability', (0, 0))\n",
    "        device = details.get('device_name', '')\n",
    "\n",
    "        print(f'Using GPU: {device}, compute capability {cc[0]}.{cc[1]}')\n",
    "        print('Memory growth enabled')\n",
    "\n",
    "    except RuntimeError as e:\n",
    "        print(e)\n",
    "\n",
    "else:\n",
    "    print('No GPU found')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f5271c9d",
   "metadata": {},
   "source": [
    "## 1. Data preparation\n",
    "\n",
    "### 1.1. Load occupancy data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0776b899",
   "metadata": {},
   "outputs": [],
   "source": [
    "occupancy_df = pd.read_csv('https://gperdrizet.github.io/FSA_devops/assets/data/unit4/occupancy_data.csv')\n",
    "occupancy_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "388df321",
   "metadata": {},
   "outputs": [],
   "source": [
    "occupancy_df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "42232b35",
   "metadata": {},
   "outputs": [],
   "source": [
    "label = 'Occupancy'\n",
    "features = ['Temperature','Humidity','Light','CO2','HumidityRatio']"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cae08764",
   "metadata": {},
   "source": [
    "### 1.2. Train test split\n",
    "\n",
    "Split the data into training and testing sets. Remember to set a random state for reproducibility."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a04f9e25",
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: Split the data"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e80d04ed",
   "metadata": {},
   "source": [
    "### 1.3. Standard scale\n",
    "\n",
    "Scale the features to improve model performance. Consider which data should be used for fitting the scaler."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d835cc05",
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: Scale the features"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d2346c36",
   "metadata": {},
   "source": [
    "## 2. Logistic regression baseline\n",
    "\n",
    "Create a baseline model to compare against the neural network.\n",
    "\n",
    "### 2.1. Fit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7e2635d5",
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: Create and fit the model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "42b30c63",
   "metadata": {},
   "source": [
    "### 2.2. Test set evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "42f41ab0",
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: Evaluate on test set"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2bdbd84c",
   "metadata": {},
   "source": [
    "## 3. Keras Sequential API model\n",
    "\n",
    "The Sequential API is the simplest way to build a neural network in Keras. It allows you to create models layer-by-layer in a linear stack.\n",
    "\n",
    "### 3.1. Build model\n",
    "\n",
    "Build a sequential neural network with an appropriate architecture for binary classification. Consider:\n",
    "- What input shape is needed based on the number of features\n",
    "- How many hidden layers and neurons per layer\n",
    "- What activation functions to use\n",
    "- The output layer configuration for binary classification\n",
    "- The optimizer, loss function, and metrics for compilation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8d0511fe",
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: Build and compile the model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7edbedf4",
   "metadata": {},
   "source": [
    "### 3.2. Train model\n",
    "\n",
    "Train the model on your prepared data. You'll need to specify the number of epochs, batch size, and validation strategy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "158b7c29",
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: Train the model and review final metrics"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e855aa43",
   "metadata": {},
   "source": [
    "### 3.3. Learning curves\n",
    "\n",
    "Visualize how the model's loss and accuracy changed during training."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "79038e76",
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: Plot learning curves"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b9edb80f",
   "metadata": {},
   "source": [
    "### 3.4. Test set evaluation\n",
    "\n",
    "Evaluate your neural network on the test set. Remember that the model outputs probabilities that need to be converted to class predictions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ca8f111b",
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: Make predictions and calculate accuracy"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e36bf90d",
   "metadata": {},
   "source": [
    "### 3.5. Performance analysis\n",
    "\n",
    "Create a confusion matrix to better understand model performance on different classes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b4d06622",
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: Display confusion matrix"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "13a53d68",
   "metadata": {},
   "source": [
    "## 4. Model comparison\n",
    "\n",
    "Compare the performance of your logistic regression baseline with the neural network."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "61f7c586",
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: Compare model accuracies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3bc8200f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# TODO: Create side-by-side confusion matrix visualization"
   ]
  }
 ],
 "metadata": {
  "language_info": {
   "name": "python"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}