TD2_Machine_Learning_correction.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Introduction to Machine-Learning Practical #2\n",
    "**By Corentin Meyer, PhD Student @ CSTB - iCube, 10/01 and 11/01 at the ESBS**  \n",
    "In this practical I will try to show you how to create a machine-learning model to predict the diabetes risk of a person.  \n",
    "We will perform data exploration, data processing, model training and evaluation and finally, fine-tuning.  \n",
    "At different parts of this practical I will add screenshots of the lecture that I did at the beginning of the year to remind you of the concepts.  \n",
    "You can access the lecture PDF here: https://raw.githubusercontent.com/lambda-science/machine-learning-introduction/main/CM_Intro_ML_Corentin_Meyer.pdf  \n",
    "\n",
    "### Context\n",
    "We will use the Diabetes dataset, a survey of more than 70,000 american people with 17 measurements per person and if they have diabetes or not. For the sake of this practical I reduced the dataset to ~5,500 people.  \n",
    "**Today we will try to build a machine-learning model that can predict if a person will likely develop diabetes or not.**  \n",
    "If you want more details on the data you can find  them [here](https://www.kaggle.com/datasets/alexteboul/diabetes-health-indicators-dataset)  \n",
    "\n",
    "<details><summary>We will use a total of 17 measurements to predict the diabetes risk. Click here to open the meaning of each column of the file</summary>\n",
    "<p>\n",
    " <ul>\n",
    "  <li><strong>Age:</strong> age of the person [18-95]</li>\n",
    "  <li><strong>Sex:</strong> sex of the person [male, female]</li>\n",
    "  <li><strong>High Chol:</strong> high level of cholesterol [yes, no]</li>\n",
    "  <li><strong>CholCheck:</strong> cholesterol check in last 5 years [yes, no]</li>\n",
    "  <li><strong>BMI:</strong> body mass index [15-100]</li>\n",
    "  <li><strong>Smoker:</strong> does the person smoke ? [0-1]</li>\n",
    "  <li><strong>HeartDiseaseorAttack:</strong> had heart issues [0-1]</li>\n",
    "  <li><strong>PhysActivity:</strong> physical activity past 30 days [0-1]</li>\n",
    "  <li><strong>Fruits:</strong> eat fruit every days [0-1]</li>\n",
    "  <li><strong>Veggies:</strong> eat veggies every days [0-1]</li>\n",
    "  <li><strong>HvyAlcoholConsump:</strong> heavy alcohol drinker [0-1]</li>\n",
    "  <li><strong>GenHlth:</strong> self-assessed general health level [excellent, very good, good, fair, poor]</li>\n",
    "  <li><strong>MentHlth:</strong> days of poor mental health during last month [0-30]</li>\n",
    "  <li><strong>PhysHlth:</strong> days of physical illness last 30 days [0-30]</li>\n",
    "  <li><strong>DiffWalk:</strong> difficulty walking or climbing stairs [0-1]</li>\n",
    "  <li><strong>Stroke:</strong> already had a stroke: [0-1]</li>\n",
    "  <li><strong>HighBP:</strong> high blood pressure [0-1]</li>\n",
    "</ul> \n",
    "</p>\n",
    "</details>\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--2023-02-15 15:45:09--  https://raw.githubusercontent.com/lambda-science/ML-TD-ESBS-2/main/diabetes_data_TD.csv\n",
      "Resolving raw.githubusercontent.com (raw.githubusercontent.com)... 185.199.108.133, 185.199.109.133, 185.199.111.133, ...\n",
      "Connecting to raw.githubusercontent.com (raw.githubusercontent.com)|185.199.108.133|:443... connected.\n",
      "HTTP request sent, awaiting response... 200 OK\n",
      "Length: 443937 (434K) [text/plain]\n",
      "Saving to: ‘diabetes_data_TD.csv.2’\n",
      "\n",
      "diabetes_data_TD.cs 100%[===================>] 433.53K  --.-KB/s    in 0.04s   \n",
      "\n",
      "2023-02-15 15:45:09 (10.2 MB/s) - ‘diabetes_data_TD.csv.2’ saved [443937/443937]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# With this command, you can download the data we will use on the server. Simply run the cell.\n",
    "!wget https://raw.githubusercontent.com/lambda-science/ML-TD-ESBS-2/main/diabetes_data_TD.csv\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# A tip before starting\n",
    "The most useful tip I can give your for this pratical but also for code in general is that **everytime you are lost on how to use a function or what function to use, simply google \"library_of_the_function name_of_the_function\" and click this first result**. You will almost always have an example on how to use the function and all the parameters needed.  \n",
    "For example: \n",
    "* \"scikit train_test_split\" -> https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.train_test_split.html"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Part 1 - Data Exploration\n",
    "### Tasks: explore the data\n",
    "* import the data\n",
    "* count the number of persons and visualize the 5 first line\n",
    "* see the diabetes vs non-diabetes person ratio\n",
    "* plot a box-plot of the BMI and an histogram of the smokers\n",
    "### Questions\n",
    "1. How many persons are in this dataset ? (rows). How many features/measurements ? (columns)\n",
    "2. What is the percentage of persons with diabetes ?\n",
    "3. What is the median BMI of the persons in the dataset ? (approximately with BoxPlot)\n",
    "4. Is there more non-smoker or smokers in the dataset ? (Histogram)  \n",
    "\n",
    "![image](https://i.imgur.com/xLoQ27w.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=== QUESTION 1 ===\n",
      "(5631, 18)\n"
     ]
    },
    {
     "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>Age</th>\n",
       "      <th>Sex</th>\n",
       "      <th>HighChol</th>\n",
       "      <th>CholCheck</th>\n",
       "      <th>BMI</th>\n",
       "      <th>Smoker</th>\n",
       "      <th>HeartDiseaseorAttack</th>\n",
       "      <th>PhysActivity</th>\n",
       "      <th>Fruits</th>\n",
       "      <th>Veggies</th>\n",
       "      <th>HvyAlcoholConsump</th>\n",
       "      <th>GenHlth</th>\n",
       "      <th>MentHlth</th>\n",
       "      <th>PhysHlth</th>\n",
       "      <th>DiffWalk</th>\n",
       "      <th>Stroke</th>\n",
       "      <th>HighBP</th>\n",
       "      <th>Diabetes</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>72.0</td>\n",
       "      <td>female</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>40.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>good</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>13.0</td>\n",
       "      <td>female</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>32.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>fair</td>\n",
       "      <td>0.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>67.0</td>\n",
       "      <td>female</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>20.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>very good</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>78.0</td>\n",
       "      <td>female</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>28.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>fair</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>56.0</td>\n",
       "      <td>female</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>28.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>good</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Age     Sex HighChol CholCheck   BMI  Smoker  HeartDiseaseorAttack  \\\n",
       "0  72.0  female      Yes       Yes  40.0     1.0                   1.0   \n",
       "1  13.0  female      Yes       Yes  32.0     0.0                   0.0   \n",
       "2  67.0  female       No       Yes  20.0     1.0                   0.0   \n",
       "3  78.0  female      Yes       Yes  28.0     0.0                   1.0   \n",
       "4  56.0  female      Yes       Yes  28.0     0.0                   0.0   \n",
       "\n",
       "   PhysActivity  Fruits  Veggies  HvyAlcoholConsump    GenHlth  MentHlth  \\\n",
       "0           0.0     1.0      1.0                0.0       good       0.0   \n",
       "1           1.0     1.0      1.0                0.0       fair       0.0   \n",
       "2           0.0     1.0      1.0                0.0  very good       0.0   \n",
       "3           0.0     1.0      1.0                0.0       fair       0.0   \n",
       "4           1.0     1.0      1.0                0.0       good       0.0   \n",
       "\n",
       "   PhysHlth  DiffWalk  Stroke  HighBP  Diabetes  \n",
       "0       0.0       0.0     0.0     1.0       1.0  \n",
       "1       5.0       0.0     0.0     1.0       1.0  \n",
       "2       0.0       0.0     0.0     1.0       1.0  \n",
       "3       0.0       1.0     0.0     1.0       1.0  \n",
       "4       0.0       0.0     0.0     0.0       1.0  "
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Pandas is the main library for data manipulation in Python\n",
    "import pandas as pd\n",
    "# Matplotlib is the main library for data visualization in Python\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Import the data (df variable) using pd.read_csv()\n",
    "df = pd.read_csv(\"diabetes_data_TD.csv\")\n",
    "\n",
    "print(\"=== QUESTION 1 ===\")\n",
    "# Print the shape of the dataframe and the head using df.shape and df.head()\n",
    "print(df.shape)\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=== QUESTION 2 ===\n",
      "0.0    0.917421\n",
      "1.0    0.082579\n",
      "Name: Diabetes, dtype: float64\n"
     ]
    }
   ],
   "source": [
    "print(\"=== QUESTION 2 ===\")\n",
    "# Use value_counts() on the \"stroke\" column. To select a column, use df[\"column_name\"]\n",
    "# Tip: set the \"normalize\" parameter to True in the value_count() function !\n",
    "print(df[\"Diabetes\"].value_counts(normalize=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=== QUESTION 3 ===\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(\"=== QUESTION 3 ===\")\n",
    "# Make a boxplot of the \"BMI\" column using the boxplot() function on the dataframe with the parameter \"column\"=\"column_name\"\n",
    "boxplot = df.boxplot(column=\"BMI\")\n",
    "# We use plt.show() to display the boxplot\n",
    "plt.show(boxplot)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=== QUESTION 4 ===\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(\"=== QUESTION 4 ===\")\n",
    "# Make a histogram of the \"Smoker\" column using the hist() function on the dataframe with the parameter \"column\"=\"column_name\"\n",
    "hist = df.hist(column=\"Smoker\")\n",
    "# We use plt.show() to display the histogram\n",
    "plt.show(hist)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Part 2 - Data Processing and splitting\n",
    "### Tasks: encode the data to be usable for training by a ML algorithm\n",
    "* analyze what columns we need to process and modify them to do machine-learning !\n",
    "* split our dataset between a training dataset and a test dataset\n",
    "\n",
    "### Questions\n",
    "1. What columns are categorical data, what columns are numeric.\n",
    "2. What columns are already ready to be used and needs no change.\n",
    "3. What type of processing do you need to do on categorical data and why\n",
    "4. What type of processing do you need to do on numeric data and why\n",
    "5. What columns contains missing data ? What type of processing do you need to do in this case.\n",
    "\n",
    "![image](https://i.imgur.com/HbEoLeR.png)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=== QUESTION 1 to 4 ===\n"
     ]
    },
    {
     "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>Age</th>\n",
       "      <th>Sex</th>\n",
       "      <th>HighChol</th>\n",
       "      <th>CholCheck</th>\n",
       "      <th>BMI</th>\n",
       "      <th>Smoker</th>\n",
       "      <th>HeartDiseaseorAttack</th>\n",
       "      <th>PhysActivity</th>\n",
       "      <th>Fruits</th>\n",
       "      <th>Veggies</th>\n",
       "      <th>HvyAlcoholConsump</th>\n",
       "      <th>GenHlth</th>\n",
       "      <th>MentHlth</th>\n",
       "      <th>PhysHlth</th>\n",
       "      <th>DiffWalk</th>\n",
       "      <th>Stroke</th>\n",
       "      <th>HighBP</th>\n",
       "      <th>Diabetes</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>72.0</td>\n",
       "      <td>female</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>40.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>good</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>13.0</td>\n",
       "      <td>female</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>32.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>fair</td>\n",
       "      <td>0.0</td>\n",
       "      <td>5.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>67.0</td>\n",
       "      <td>female</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "      <td>20.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>very good</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>78.0</td>\n",
       "      <td>female</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>28.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>fair</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>56.0</td>\n",
       "      <td>female</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>28.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>good</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    Age     Sex HighChol CholCheck   BMI  Smoker  HeartDiseaseorAttack  \\\n",
       "0  72.0  female      Yes       Yes  40.0     1.0                   1.0   \n",
       "1  13.0  female      Yes       Yes  32.0     0.0                   0.0   \n",
       "2  67.0  female       No       Yes  20.0     1.0                   0.0   \n",
       "3  78.0  female      Yes       Yes  28.0     0.0                   1.0   \n",
       "4  56.0  female      Yes       Yes  28.0     0.0                   0.0   \n",
       "\n",
       "   PhysActivity  Fruits  Veggies  HvyAlcoholConsump    GenHlth  MentHlth  \\\n",
       "0           0.0     1.0      1.0                0.0       good       0.0   \n",
       "1           1.0     1.0      1.0                0.0       fair       0.0   \n",
       "2           0.0     1.0      1.0                0.0  very good       0.0   \n",
       "3           0.0     1.0      1.0                0.0       fair       0.0   \n",
       "4           1.0     1.0      1.0                0.0       good       0.0   \n",
       "\n",
       "   PhysHlth  DiffWalk  Stroke  HighBP  Diabetes  \n",
       "0       0.0       0.0     0.0     1.0       1.0  \n",
       "1       5.0       0.0     0.0     1.0       1.0  \n",
       "2       0.0       0.0     0.0     1.0       1.0  \n",
       "3       0.0       1.0     0.0     1.0       1.0  \n",
       "4       0.0       0.0     0.0     0.0       1.0  "
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(\"=== QUESTION 1 to 4 ===\")\n",
    "# You should be able to answer the questions 1 to 4 simply with the table below (df.head())\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=== QUESTION 5 ===\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 5631 entries, 0 to 5630\n",
      "Data columns (total 18 columns):\n",
      " #   Column                Non-Null Count  Dtype  \n",
      "---  ------                --------------  -----  \n",
      " 0   Age                   5631 non-null   float64\n",
      " 1   Sex                   5631 non-null   object \n",
      " 2   HighChol              5631 non-null   object \n",
      " 3   CholCheck             5631 non-null   object \n",
      " 4   BMI                   5233 non-null   float64\n",
      " 5   Smoker                5328 non-null   float64\n",
      " 6   HeartDiseaseorAttack  5631 non-null   float64\n",
      " 7   PhysActivity          5631 non-null   float64\n",
      " 8   Fruits                5631 non-null   float64\n",
      " 9   Veggies               5631 non-null   float64\n",
      " 10  HvyAlcoholConsump     5631 non-null   float64\n",
      " 11  GenHlth               5631 non-null   object \n",
      " 12  MentHlth              5631 non-null   float64\n",
      " 13  PhysHlth              5631 non-null   float64\n",
      " 14  DiffWalk              5631 non-null   float64\n",
      " 15  Stroke                5062 non-null   float64\n",
      " 16  HighBP                5631 non-null   float64\n",
      " 17  Diabetes              5631 non-null   float64\n",
      "dtypes: float64(14), object(4)\n",
      "memory usage: 792.0+ KB\n"
     ]
    }
   ],
   "source": [
    "print(\"=== QUESTION 5 ===\")\n",
    "# You should be able to answer the questions 5 simply with the table below (df.info())\n",
    "# (remember how many rows there are in the dataframe)\n",
    "df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "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>Sex</th>\n",
       "      <th>HighChol</th>\n",
       "      <th>CholCheck</th>\n",
       "      <th>GenHlth</th>\n",
       "      <th>Age</th>\n",
       "      <th>BMI</th>\n",
       "      <th>MentHlth</th>\n",
       "      <th>PhysHlth</th>\n",
       "      <th>Smoker</th>\n",
       "      <th>HeartDiseaseorAttack</th>\n",
       "      <th>PhysActivity</th>\n",
       "      <th>Fruits</th>\n",
       "      <th>Veggies</th>\n",
       "      <th>HvyAlcoholConsump</th>\n",
       "      <th>DiffWalk</th>\n",
       "      <th>Stroke</th>\n",
       "      <th>HighBP</th>\n",
       "      <th>Diabetes</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>1.146139</td>\n",
       "      <td>1.902009</td>\n",
       "      <td>-0.426205</td>\n",
       "      <td>-0.470133</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>-2.179834</td>\n",
       "      <td>0.622491</td>\n",
       "      <td>-0.426205</td>\n",
       "      <td>0.124957</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>4.0</td>\n",
       "      <td>0.864277</td>\n",
       "      <td>-1.296787</td>\n",
       "      <td>-0.426205</td>\n",
       "      <td>-0.470133</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.484374</td>\n",
       "      <td>-0.017268</td>\n",
       "      <td>-0.426205</td>\n",
       "      <td>-0.470133</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>0.244180</td>\n",
       "      <td>-0.017268</td>\n",
       "      <td>-0.426205</td>\n",
       "      <td>-0.470133</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Sex  HighChol  CholCheck  GenHlth       Age       BMI  MentHlth  PhysHlth  \\\n",
       "0  0.0       1.0        1.0      2.0  1.146139  1.902009 -0.426205 -0.470133   \n",
       "1  0.0       1.0        1.0      1.0 -2.179834  0.622491 -0.426205  0.124957   \n",
       "2  0.0       0.0        1.0      4.0  0.864277 -1.296787 -0.426205 -0.470133   \n",
       "3  0.0       1.0        1.0      1.0  1.484374 -0.017268 -0.426205 -0.470133   \n",
       "4  0.0       1.0        1.0      2.0  0.244180 -0.017268 -0.426205 -0.470133   \n",
       "\n",
       "   Smoker  HeartDiseaseorAttack  PhysActivity  Fruits  Veggies  \\\n",
       "0     1.0                   1.0           0.0     1.0      1.0   \n",
       "1     0.0                   0.0           1.0     1.0      1.0   \n",
       "2     1.0                   0.0           0.0     1.0      1.0   \n",
       "3     0.0                   1.0           0.0     1.0      1.0   \n",
       "4     0.0                   0.0           1.0     1.0      1.0   \n",
       "\n",
       "   HvyAlcoholConsump  DiffWalk  Stroke  HighBP  Diabetes  \n",
       "0                0.0       0.0     0.0     1.0       1.0  \n",
       "1                0.0       0.0     0.0     1.0       1.0  \n",
       "2                0.0       0.0     0.0     1.0       1.0  \n",
       "3                0.0       1.0     0.0     1.0       1.0  \n",
       "4                0.0       0.0     0.0     0.0       1.0  "
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# We will now process theses columns to make them usable\n",
    "# From Scikit learn we import the tools we will need\n",
    "# OrdinalEncoder for the categorical variables\n",
    "# StandardScaler for the numerical variables\n",
    "# IterativeImputer for the missing values\n",
    "from sklearn.preprocessing import OrdinalEncoder, StandardScaler\n",
    "from sklearn.experimental import enable_iterative_imputer\n",
    "from sklearn.impute import IterativeImputer\n",
    "# NumPy is the main library for array manipulation in Python\n",
    "import numpy as np\n",
    "\n",
    "# Create lists with columns names for each processing categories\n",
    "columns_nothing_to_do = [\"Smoker\",\"HeartDiseaseorAttack\", \"PhysActivity\", \"Fruits\", \"Veggies\", \"HvyAlcoholConsump\", \"DiffWalk\", \"Stroke\", \"HighBP\", \"Diabetes\"] # 10 Columns\n",
    "categorical_columns_to_process = [\"Sex\", \"HighChol\", \"CholCheck\", \"GenHlth\"] # 4 Columns\n",
    "numerical_columns_to_process = [\"Age\", \"BMI\", \"MentHlth\", \"PhysHlth\"] # 4 Columns\n",
    "\n",
    "\n",
    "# Categorical data to numeric: do one hot encoding ( OrdinalEncoder(), .fit_transform() )\n",
    "# I give you the OrdinalEncoder example. Do something similar for the Scaling of numerics data !\n",
    "X_cat = df[categorical_columns_to_process]\n",
    "X_cat_enc = OrdinalEncoder().fit_transform(X_cat)\n",
    "\n",
    "# Numeric data: do scaling (-1,+1) (StandardScaler() and .fit_transform())\n",
    "X_num = df[numerical_columns_to_process]\n",
    "X_num_scaled = StandardScaler().fit_transform(X_num)\n",
    "\n",
    "# Nothing-To-Do Cols: just select the data corresponding with df[name_of_the_list] and store it into X_nothing_to_do\n",
    "# Then convert X_nothing_to_do to numpy array with .to_numpy()\n",
    "X_nothing_to_do = df[columns_nothing_to_do]\n",
    "X_nothing_to_do = X_nothing_to_do.to_numpy()\n",
    "\n",
    "# Combine the X_cat_enc, X_num_scaled, and X_nothing_to_do data into a single array (array_data)\n",
    "# using np.concatenate((array1, array2, array3), parameters) ; TIP: Care for the axis parameter (axis=1)\n",
    "array_data = np.concatenate((X_cat_enc, X_num_scaled, X_nothing_to_do), axis=1)\n",
    "\n",
    "# Finally handle Missing Data: Use imputer to predict them ( IterativeImputer() and .fit_transform() on array_data just as before\n",
    "array_data = IterativeImputer().fit_transform(array_data)\n",
    "\n",
    "# You can recreate a DataFrame for pretty printing, but this is optional. I give you the code it's a bit ugly.\n",
    "df_data = pd.DataFrame(data=array_data, columns=list(X_cat.columns) + list(X_num.columns) + list(df[columns_nothing_to_do].columns))\n",
    "df_data.head()"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Tasks: Split our dataset between train and test set\n",
    "* split our processed data into train and test dataset\n",
    "\n",
    "### Questions\n",
    "1. What train/test ratio should you use.\n",
    "2. How many entries are in your train dataset and in your test dataset.\n",
    "3. Verify that you have the same diabetes / diabetes ratio between train and test dataset.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4223, 17)\n",
      "(1408, 17)\n",
      "[3885  338]\n",
      "[1281  127]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "# From your numpy array that you created, separate the X columns (features: multiples columns) and the Y column (label: last column) in two variables\n",
    "# Call them X and Y, X contains all the columns except the last one, Y contains only the last column from the array_data variable.\n",
    "X = array_data[:,:-1]\n",
    "Y = array_data[:,-1]\n",
    "\n",
    "# Use train_test_split() using X and Y with the ratio you selected. The parameter test_size= is the ratio of the test set.\n",
    "X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.25, random_state=777)\n",
    "\n",
    "# Print X_train and X_test shape, use np.unique() on Y_train and Y_test to get the diab vs no diab ratio\n",
    "print(X_train.shape)\n",
    "print(X_test.shape)\n",
    "print(np.unique(Y_train, return_counts=True)[1])\n",
    "print(np.unique(Y_test, return_counts=True)[1])"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Part 3 - Create your machine-learning model\n",
    "### Tasks: Choose a model and do basic evaluation\n",
    "\n",
    "Select from scikit-learn a model and train it (.fit()) with the train data. Then calculate the accuracy of the model on the test data (.score()). Plot the confusion matrix of the test data classification.  \n",
    "All ML algorithm in Sci-kit can be found here (1.1 to 1.11) https://scikit-learn.org/stable/supervised_learning.html  \n",
    "Personally I will use a RandomForest or a Decision Tree for example.\n",
    "\n",
    "### Questions:\n",
    "\n",
    "1. Which model did you choose and why ? Have you set any particular (hyper)parameters ?\n",
    "2. What accuracy-score do you get and what conclusion can you take ?\n",
    "3. What do you observe on the confusion matrix and what conclusion can you take ?\n",
    "\n",
    "![image](https://i.imgur.com/JttOOSw.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=== QUESTION 2-3 ===\n",
      "0.9090909090909091\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAhEAAAGwCAYAAAAXNjfEAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy88F64QAAAACXBIWXMAAA9hAAAPYQGoP6dpAAA8bElEQVR4nO3de1yUdfr/8fcAchAZEEsQRcM0D2VaWkYH02+saG7pN/u2FhWZ6VZSHspTq2ZaUXbWTDtq7k83bXdzy1yL1TwlaqK0akiZFqaB7SIgGKeZ+/eHMTXJPTEHBMbX8/G4H7tz35/7nmtmWbm4rs/9uS2GYRgCAABwU0BDBwAAAJomkggAAOARkggAAOARkggAAOARkggAAOARkggAAOARkggAAOCRoIYO4Eyz2+06evSoIiIiZLFYGjocAICbDMPQiRMnFBcXp4CA+vtbuLy8XJWVlV5fJzg4WKGhoT6IqPE565KIo0ePKj4+vqHDAAB46fDhw2rXrl29XLu8vFwJHVoo/5jN62vFxsbq0KFDfplInHVJREREhCTp213nydqCbg780/9e0KOhQwDqTbWqtEVrHP+e14fKykrlH7Pp26zzZI3w/HdFyQm7OvT+RpWVlSQR/qCmhWFtEeDVDwbQmAVZmjV0CED9+elhDWeiJd0iwqIWEZ6/j13+3TY/65IIAADqymbYZfPiCVM2w+67YBohkggAAEzYZcguz7MIb85tCqjnAwAAj1CJAADAhF12edOQ8O7sxo8kAgAAEzbDkM3wvCXhzblNAe0MAADgESoRAACYYGKlayQRAACYsMuQjSTCFO0MAADgESoRAACYoJ3hGkkEAAAmuDvDNdoZAADAI1QiAAAwYf9p8+Z8f0YSAQCACZuXd2d4c25TQBIBAIAJmyEvn+Lpu1gaI+ZEAAAAj5BEAABgwu6DzR2bNm3SDTfcoLi4OFksFq1atcpxrKqqSlOmTFGPHj0UHh6uuLg43XnnnTp69KjTNQoLC5WSkiKr1aqoqCiNGjVKpaWlTmP+/e9/65prrlFoaKji4+M1d+5cNyM9hSQCAAATdllk82Kzy+LW+5WVlalnz55asGDBacdOnjypXbt2acaMGdq1a5f+/ve/Kzc3VzfeeKPTuJSUFO3bt08ZGRlavXq1Nm3apDFjxjiOl5SUaODAgerQoYOysrL0zDPPaNasWXrttdfc/n6YEwEAQD0rKSlxeh0SEqKQkJDTxg0ePFiDBw+u9RqRkZHKyMhw2vfyyy/r8ssvV15entq3b6+cnBytXbtWn332mfr06SNJmj9/vq6//no9++yziouL07Jly1RZWam33npLwcHBuvDCC5Wdna3nn3/eKdmoCyoRAACYsBveb5IUHx+vyMhIx5aenu6T+IqLi2WxWBQVFSVJyszMVFRUlCOBkKSkpCQFBARo+/btjjH9+vVTcHCwY0xycrJyc3N1/Phxt96fSgQAACZq2hLenC9Jhw8fltVqdeyvrQrhrvLyck2ZMkW33nqr49r5+flq3bq107igoCBFR0crPz/fMSYhIcFpTExMjONYy5Yt6xwDSQQAAPXMarU6JRHeqqqq0i233CLDMLRw4UKfXdddJBEAAJjwVSXCl2oSiG+//Vbr1693Sk5iY2N17Ngxp/HV1dUqLCxUbGysY0xBQYHTmJrXNWPqijkRAACYsBsWrzdfqkkgvvrqK/3rX/9Sq1atnI4nJiaqqKhIWVlZjn3r16+X3W5X3759HWM2bdqkqqoqx5iMjAx16dLFrVaGRBIBAECjUVpaquzsbGVnZ0uSDh06pOzsbOXl5amqqko333yzdu7cqWXLlslmsyk/P1/5+fmqrKyUJHXr1k2DBg3S6NGjtWPHDn366adKS0vTiBEjFBcXJ0m67bbbFBwcrFGjRmnfvn1asWKFXnrpJU2cONHteGlnAABg4ky3M3bu3KkBAwY4Xtf8Yk9NTdWsWbP0/vvvS5J69erldN4nn3yi/v37S5KWLVumtLQ0XXfddQoICNDw4cM1b948x9jIyEh9/PHHGjt2rHr37q1zzjlHM2fOdPv2TokkAgAAUzYFyOZF0d7m5vj+/fvLMMwfuOHqWI3o6GgtX77c5ZiLL75YmzdvdjO605FEAABgwvByXoPh4zkRjQ1zIgAAgEeoRAAAYKIx3uLZmJBEAABgwmYEyGZ4MSfit6cwNGm0MwAAgEeoRAAAYMIui+xe/L1tl3+XIkgiAAAwwZwI12hnAAAAj1CJAADAhPcTK2lnAABwVjo1J8LzloQ35zYFtDMAAIBHqEQAAGDC7uWzM7g7AwCAsxRzIlwjiQAAwIRdAawT4QJzIgAAgEeoRAAAYMJmWGTz4nHe3pzbFJBEAABgwublxEob7QwAAIDTUYkAAMCE3QiQ3Yu7M+zcnQEAwNmJdoZrtDMAAIBHqEQAAGDCLu/usLD7LpRGiSQCAAAT3i825d8Ff//+dAAAoN5QiQAAwIT3z87w77/VSSIAADBhl0V2eTMnghUrAQA4K1GJcM2/Px0AAKg3VCIAADDh/WJT/v23OkkEAAAm7IZFdm/WifDzp3j6d4oEAADqDZUIAABM2L1sZ/j7YlMkEQAAmPD+KZ7+nUT496cDAAD1hkoEAAAmbLLI5sWCUd6c2xSQRAAAYIJ2hmv+/ekAAEC9oRIBAIAJm7xrSdh8F0qjRBIBAIAJ2hmukUQAAGCCB3C55t+fDgAA1BsqEQAAmDBkkd2LOREGt3gCAHB2op3hmn9/OgAAUG+oRAAAYIJHgbtGEgEAgAmbl0/x9ObcpsC/Px0AAKg3JBEAAJioaWd4s7lj06ZNuuGGGxQXFyeLxaJVq1Y5HTcMQzNnzlSbNm0UFhampKQkffXVV05jCgsLlZKSIqvVqqioKI0aNUqlpaVOY/7973/rmmuuUWhoqOLj4zV37lyPvh+SCAAATNgV4PXmjrKyMvXs2VMLFiyo9fjcuXM1b948LVq0SNu3b1d4eLiSk5NVXl7uGJOSkqJ9+/YpIyNDq1ev1qZNmzRmzBjH8ZKSEg0cOFAdOnRQVlaWnnnmGc2aNUuvvfaa298PcyIAAKhnJSUlTq9DQkIUEhJy2rjBgwdr8ODBtV7DMAy9+OKLmj59uoYOHSpJWrp0qWJiYrRq1SqNGDFCOTk5Wrt2rT777DP16dNHkjR//nxdf/31evbZZxUXF6dly5apsrJSb731loKDg3XhhRcqOztbzz//vFOyURdUIgAAMGEzLF5vkhQfH6/IyEjHlp6e7nYshw4dUn5+vpKSkhz7IiMj1bdvX2VmZkqSMjMzFRUV5UggJCkpKUkBAQHavn27Y0y/fv0UHBzsGJOcnKzc3FwdP37crZioRAAAYMJXt3gePnxYVqvVsb+2KsRvyc/PlyTFxMQ47Y+JiXEcy8/PV+vWrZ2OBwUFKTo62mlMQkLCadeoOdayZcs6x0QSAQCACcPLp3gaP51rtVqdkgh/QTsDAIAmIDY2VpJUUFDgtL+goMBxLDY2VseOHXM6Xl1drcLCQqcxtV3jl+9RVyQRAACYsMni9eYrCQkJio2N1bp16xz7SkpKtH37diUmJkqSEhMTVVRUpKysLMeY9evXy263q2/fvo4xmzZtUlVVlWNMRkaGunTp4lYrQyKJAADAlN3wdq0I996vtLRU2dnZys7OlnRqMmV2drby8vJksVg0fvx4Pf7443r//fe1Z88e3XnnnYqLi9OwYcMkSd26ddOgQYM0evRo7dixQ59++qnS0tI0YsQIxcXFSZJuu+02BQcHa9SoUdq3b59WrFihl156SRMnTnT7+2FOBAAAjcTOnTs1YMAAx+uaX+ypqalasmSJJk+erLKyMo0ZM0ZFRUW6+uqrtXbtWoWGhjrOWbZsmdLS0nTdddcpICBAw4cP17x58xzHIyMj9fHHH2vs2LHq3bu3zjnnHM2cOdPt2zslyWIYhpt5UtNWUlKiyMhIHf+yo6wRFGLqas+2cL37Smt9tae5Cgua6dE3D+nKwcWSpOoqacnTbfTZequ+/zZY4Va7LrnmhEY9clStYqslSZ9vbaHJN3eq9drz1uSqS68fJUmGIf110bn657JWOvZdsKzR1fp96n9127iCWs9F7ZLjejV0CH7v9ofydcdDzj+Xhw+E6J5+XRsoorNHtVGlDfqHiouL622yYs3vitRPRii4RfBvn2CisrRSbw94p15jbUhUIlAn5ScD1PHCH5V8a6Fmj3K+NajixwAd2NNct40vUMfuP6q0OFALZ7bVo3d11Mtrv5Qkde9Tpr9k73U67+25bZS9pYUu6PmjY9/CGW2VtTFCo2ccVUK3cp0oClTJ8cD6/4CAB77ZH6qpf+joeG2z+fcTG89Gdllk92JegzfnNgWN4k/xBQsW6LzzzlNoaKj69u2rHTt2uBz/7rvvqmvXrgoNDVWPHj20Zs2aMxTp2euy/zmhu6bk66qfqg+/FG6166kVX+vaG4sU36lC3Xqf1NgnvtNX/26uY981kyQ1CzYU3brasVlbVivzI6sG/qFQlp/+P5b3VYhWLz1HsxYfUmJyiWLbV6rzxT+q97Wlp70n0BjYbNLxH5o5tpJC/i7D2aXBk4gVK1Zo4sSJevTRR7Vr1y717NlTycnJp92iUmPr1q269dZbNWrUKO3evVvDhg3TsGHDtHfv3lrHo2GUlQTKYjEUHmmr9Xjmx5E6cTxIA/9Q6Ni37eNItWlfoe3/surOvt105+Xd9cJD8VQi0Gi1TajU8l37tCQzR1Ne/lbntq1s6JDgY75asdJfNXgS8fzzz2v06NEaOXKkunfvrkWLFql58+Z66623ah3/0ksvadCgQZo0aZK6deumOXPm6NJLL9XLL798hiOHmcpyi958Ik79hx1XeIS91jEf/aWVevc/oXPjfr7F6Pu8YBUcCdbm1VGaNC9PD72Yp6/+HabHx5x3hiIH6m7/ruZ6dny8/pTSUfOntlVs+0o9994BhYXXnjijabL/tNiUN5s/a9BPV1lZqaysLKd1wAMCApSUlORYB/zXMjMzncZLp9b8NhtfUVGhkpISpw31p7pKeuKP50mG9MBT39U65oejzZS1IULJt/7Xab9hl6oqAjTppTz16FumnleWasJzh/X5pxE6fMD9JWKB+rTzE6s2r47SoZwwZW20avrtHdXCalO/G4saOjTgjGnQJOI///mPbDaby3XAfy0/P9+t8enp6U4PPYmPj/dN8DhNTQJRcCRY6e98bVqF+HhFtCJaVitxoPP8iujW1QoMMtTu/ArHvvadTz3e9tiRZvUXOOADZSWB+u5giOLOo6XhT+zyZo0I7yZlNgX+XWeRNG3aNBUXFzu2w4cPN3RIfqkmgThyKERPrTgga3TtJV3DOJVEJN18XEG/ygsuvKxMtmqLjn7z8+1U3x08VYGIaVcloDELbW5TXIdKFR5jcqU/MX66O8PTzfDzJKJBf9rPOeccBQYGulwH/NfM1vw2G2/2zHa458eyAB099PP3mH84WF/vDVNEVLWiY6o0Z3SCDuwJ0+ylB2W3WRz/kEZE2dQs+OelSLK3tFB+XogG3fbf097jkn4n1KnHST0/sb3ufeyIDEN6+ZF2urRfiVN1AmgMRs88qm0fW3Xsu2C1iq3SHQ/ny2aXNrzn3rLBaNx89RRPf9WgSURwcLB69+6tdevWOZbstNvtWrdundLS0mo9JzExUevWrdP48eMd+zIyMhzrhqN+fPl5c6fFol6d1VaS9LtbCnX7Q/na9nGkJOn+3zkvtDP3rwfU88qfb9Fc+5dW6t6nVO07n54UBARIs98+qAXT2+nhmzoptLldfQaUaMyjR+vjIwFeOadNlaa98q0iWtpU/N8g7fssXON/31nF3OaJs0iD/7RPnDhRqamp6tOnjy6//HK9+OKLKisr08iRIyVJd955p9q2bav09HRJ0rhx43Tttdfqueee05AhQ/TOO+9o586deu211xryY/i9nleW6qOj2abHXR37pWmvfOvyeKvYas1845u6BwY0kPT7OjR0CDgDvL3Dwt/vzmjwJOIPf/iDfvjhB82cOVP5+fnq1auX1q5d65g8mZeXp4CAn/9HuPLKK7V8+XJNnz5djzzyiDp37qxVq1bpoosuaqiPAADwU7QzXOPZGYAf4tkZ8Gdn8tkZQz++W83CPX92RlVZpf4x8C2enQEAwNmGZ2e4RhIBAIAJ2hmuUc8HAAAeoRIBAIAJKhGukUQAAGCCJMI12hkAAMAjVCIAADBBJcI1kggAAEwY8u42TX9fiIkkAgAAE1QiXGNOBAAA8AiVCAAATFCJcI0kAgAAEyQRrtHOAAAAHqESAQCACSoRrpFEAABgwjAsMrxIBLw5tymgnQEAADxCJQIAABN2WbxabMqbc5sCkggAAEwwJ8I12hkAAMAjVCIAADDBxErXSCIAADBBO8M1kggAAExQiXCNOREAAMAjVCIAADBheNnO8PdKBEkEAAAmDEmG4d35/ox2BgAA8AiVCAAATNhlkYUVK02RRAAAYIK7M1yjnQEAADxCJQIAABN2wyILi02ZIokAAMCEYXh5d4af355BOwMAAHiESgQAACaYWOkaSQQAACZIIlwjiQAAwAQTK11jTgQAAI2EzWbTjBkzlJCQoLCwMJ1//vmaM2eOjF/M0DQMQzNnzlSbNm0UFhampKQkffXVV07XKSwsVEpKiqxWq6KiojRq1CiVlpb6PF6SCAAATNTcneHN5o6nn35aCxcu1Msvv6ycnBw9/fTTmjt3rubPn+8YM3fuXM2bN0+LFi3S9u3bFR4eruTkZJWXlzvGpKSkaN++fcrIyNDq1au1adMmjRkzxldfiwPtDAAATJxKBLyZE3HqP0tKSpz2h4SEKCQk5LTxW7du1dChQzVkyBBJ0nnnnae//OUv2rFjx0/XM/Tiiy9q+vTpGjp0qCRp6dKliomJ0apVqzRixAjl5ORo7dq1+uyzz9SnTx9J0vz583X99dfr2WefVVxcnMef59eoRAAAUM/i4+MVGRnp2NLT02sdd+WVV2rdunX68ssvJUmff/65tmzZosGDB0uSDh06pPz8fCUlJTnOiYyMVN++fZWZmSlJyszMVFRUlCOBkKSkpCQFBARo+/btPv1cVCIAADDhq7szDh8+LKvV6thfWxVCkqZOnaqSkhJ17dpVgYGBstlseuKJJ5SSkiJJys/PlyTFxMQ4nRcTE+M4lp+fr9atWzsdDwoKUnR0tGOMr5BEAABgwvhp8+Z8SbJarU5JhJmVK1dq2bJlWr58uS688EJlZ2dr/PjxiouLU2pqqheR1A+SCAAAGolJkyZp6tSpGjFihCSpR48e+vbbb5Wenq7U1FTFxsZKkgoKCtSmTRvHeQUFBerVq5ckKTY2VseOHXO6bnV1tQoLCx3n+wpzIgAAMFHTzvBmc8fJkycVEOD8qzkwMFB2u12SlJCQoNjYWK1bt85xvKSkRNu3b1diYqIkKTExUUVFRcrKynKMWb9+vex2u/r27evpV1ErKhEAAJjxVT+jjm644QY98cQTat++vS688ELt3r1bzz//vO6++25JksVi0fjx4/X444+rc+fOSkhI0IwZMxQXF6dhw4ZJkrp166ZBgwZp9OjRWrRokaqqqpSWlqYRI0b49M4MiSQCAABzXk6slJvnzp8/XzNmzND999+vY8eOKS4uTn/84x81c+ZMx5jJkyerrKxMY8aMUVFRka6++mqtXbtWoaGhjjHLli1TWlqarrvuOgUEBGj48OGaN2+e55/DhMUw/P1Bpc5KSkoUGRmp4192lDWCbg78U3Jcr4YOAag31UaVNugfKi4urtNkRU/U/K7ouORPCmge+tsnmLCfLNfBu56o11gbEpUIAABMeLLq5K/P92ckEQAAmOApnq5RzwcAAB6hEgEAgBnD4vbkyNPO92MkEQAAmGBOhGu0MwAAgEeoRAAAYOYMLzbV1JBEAABggrszXKtTEvH+++/X+YI33nijx8EAAICmo05JRM163L/FYrHIZrN5Ew8AAI2Ln7ckvFGnJKLm6WEAAJxNaGe45tXdGeXl5b6KAwCAxsfwwebH3E4ibDab5syZo7Zt26pFixY6ePCgJGnGjBl68803fR4gAABonNxOIp544gktWbJEc+fOVXBwsGP/RRddpDfeeMOnwQEA0LAsPtj8l9tJxNKlS/Xaa68pJSVFgYGBjv09e/bU/v37fRocAAANinaGS24nEUeOHFGnTp1O22+321VVVeWToAAAQOPndhLRvXt3bd68+bT9f/3rX3XJJZf4JCgAABoFKhEuub1i5cyZM5WamqojR47Ibrfr73//u3Jzc7V06VKtXr26PmIEAKBh8BRPl9yuRAwdOlQffPCB/vWvfyk8PFwzZ85UTk6OPvjgA/3ud7+rjxgBAEAj5NGzM6655hplZGT4OhYAABoVHgXumscP4Nq5c6dycnIknZon0bt3b58FBQBAo8BTPF1yO4n47rvvdOutt+rTTz9VVFSUJKmoqEhXXnml3nnnHbVr187XMQIAgEbI7TkR99xzj6qqqpSTk6PCwkIVFhYqJydHdrtd99xzT33ECABAw6iZWOnN5sfcrkRs3LhRW7duVZcuXRz7unTpovnz5+uaa67xaXAAADQki3Fq8+Z8f+Z2EhEfH1/rolI2m01xcXE+CQoAgEaBOREuud3OeOaZZ/TAAw9o586djn07d+7UuHHj9Oyzz/o0OAAA0HjVqRLRsmVLWSw/93XKysrUt29fBQWdOr26ulpBQUG6++67NWzYsHoJFACAM47FplyqUxLx4osv1nMYAAA0QrQzXKpTEpGamlrfcQAAgCbG48WmJKm8vFyVlZVO+6xWq1cBAQDQaFCJcMntiZVlZWVKS0tT69atFR4erpYtWzptAAD4DZ7i6ZLbScTkyZO1fv16LVy4UCEhIXrjjTf02GOPKS4uTkuXLq2PGAEAQCPkdjvjgw8+0NKlS9W/f3+NHDlS11xzjTp16qQOHTpo2bJlSklJqY84AQA487g7wyW3KxGFhYXq2LGjpFPzHwoLCyVJV199tTZt2uTb6AAAaEA1K1Z6s/kzt5OIjh076tChQ5Kkrl27auXKlZJOVShqHsgFAAD8n9tJxMiRI/X5559LkqZOnaoFCxYoNDRUEyZM0KRJk3weIAAADYaJlS65PSdiwoQJjv+elJSk/fv3KysrS506ddLFF1/s0+AAAEDj5dU6EZLUoUMHdejQwRexAADQqFjk5VM8fRZJ41SnJGLevHl1vuCDDz7ocTAAAKDpqFMS8cILL9TpYhaLpckkEf/3u8EKCghp6DCAepLX0AEA/oFbPF2qUxJRczcGAABnFZa9dsntuzMAAAAkH0ysBADAb1GJcIkkAgAAE96uOsmKlQAAALWgEgEAgBnaGS55VInYvHmzbr/9diUmJurIkSOSpD//+c/asmWLT4MDAKBBsey1S24nEX/729+UnJyssLAw7d69WxUVFZKk4uJiPfnkkz4PEACAs8mRI0d0++23q1WrVgoLC1OPHj20c+dOx3HDMDRz5ky1adNGYWFhSkpK0ldffeV0jcLCQqWkpMhqtSoqKkqjRo1SaWmpz2N1O4l4/PHHtWjRIr3++utq1qyZY/9VV12lXbt2+TQ4AAAa0pl+FPjx48d11VVXqVmzZvrnP/+pL774Qs8995xatmzpGDN37lzNmzdPixYt0vbt2xUeHq7k5GSVl5c7xqSkpGjfvn3KyMjQ6tWrtWnTJo0ZM8ZXX4uD23MicnNz1a9fv9P2R0ZGqqioyBcxAQDQOJzhFSuffvppxcfHa/HixY59CQkJP1/OMPTiiy9q+vTpGjp0qCRp6dKliomJ0apVqzRixAjl5ORo7dq1+uyzz9SnTx9J0vz583X99dfr2WefVVxcnOef51fcrkTExsbqwIEDp+3fsmWLOnbs6JOgAABoFHw0J6KkpMRpq5kK8Gvvv/+++vTpo//7v/9T69atdckll+j11193HD906JDy8/OVlJTk2BcZGam+ffsqMzNTkpSZmamoqChHAiGdeup2QECAtm/f7oMv5WduJxGjR4/WuHHjtH37dlksFh09elTLli3Tww8/rPvuu8+nwQEA4A/i4+MVGRnp2NLT02sdd/DgQS1cuFCdO3fWRx99pPvuu08PPvig3n77bUlSfn6+JCkmJsbpvJiYGMex/Px8tW7d2ul4UFCQoqOjHWN8xe12xtSpU2W323Xdddfp5MmT6tevn0JCQvTwww/rgQce8GlwAAA0JF8tNnX48GFZrVbH/pCQ2h8Aabfb1adPH8eNCpdccon27t2rRYsWKTU11fNA6onblQiLxaI//elPKiws1N69e7Vt2zb98MMPmjNnTn3EBwBAw/FRO8NqtTptZklEmzZt1L17d6d93bp1U17eqSfzxsbGSpIKCgqcxhQUFDiOxcbG6tixY07Hq6urVVhY6BjjKx6vWBkcHKzu3bvr8ssvV4sWLXwZEwAAZ6WrrrpKubm5Tvu+/PJLdejQQdKpSZaxsbFat26d43hJSYm2b9+uxMRESVJiYqKKioqUlZXlGLN+/XrZ7Xb17dvXp/G63c4YMGCALBbz2abr16/3KiAAABoNL9sZ7i42NWHCBF155ZV68skndcstt2jHjh167bXX9Nprr0k61Q0YP368Hn/8cXXu3FkJCQmaMWOG4uLiNGzYMEmnKheDBg3S6NGjtWjRIlVVVSktLU0jRozw6Z0ZkgdJRK9evZxeV1VVKTs7W3v37m2U/RoAADx2hpe9vuyyy/Tee+9p2rRpmj17thISEvTiiy8qJSXFMWby5MkqKyvTmDFjVFRUpKuvvlpr165VaGioY8yyZcuUlpam6667TgEBARo+fLjmzZvnxQepncUwDJ8syjlr1iyVlpbq2Wef9cXl6k1JSYkiIyOV1GGsggJq70kBTV31N3kNHQJQb6qNKm3QP1RcXOw0WdGXan5XdJz+pAJ/8cvZXbbych18/JF6jbUh+ewpnrfffrveeustX10OAICGx7MzXPLZUzwzMzOdSikAADR1vrrF01+5nUTcdNNNTq8Nw9D333+vnTt3asaMGT4LDAAANG5uJxGRkZFOrwMCAtSlSxfNnj1bAwcO9FlgAACgcXMribDZbBo5cqR69Ojh9EQxAAD80hm+O6OpcWtiZWBgoAYOHMjTOgEAZ4Uz/SjwpsbtuzMuuugiHTx4sD5iAQAATYjbScTjjz+uhx9+WKtXr9b3339/2uNNAQDwK9zeaarOcyJmz56thx56SNdff70k6cYbb3Ra/towDFksFtlsNt9HCQBAQ2BOhEt1TiIee+wx3Xvvvfrkk0/qMx4AANBE1DmJqFkd+9prr623YAAAaExYbMo1t27xdPX0TgAA/A7tDJfcSiIuuOCC30wkCgsLvQoIAAA0DW4lEY899thpK1YCAOCvaGe45lYSMWLECLVu3bq+YgEAoHGhneFSndeJYD4EAAD4JbfvzgAA4KxBJcKlOicRdru9PuMAAKDRYU6Ea24/ChwAgLMGlQiX3H52BgAAgEQlAgAAc1QiXCKJAADABHMiXKOdAQAAPEIlAgAAM7QzXCKJAADABO0M12hnAAAAj1CJAADADO0Ml0giAAAwQxLhEu0MAADgESoRAACYsPy0eXO+PyOJAADADO0Ml0giAAAwwS2erjEnAgAAeIRKBAAAZmhnuEQSAQCAK36eCHiDdgYAAPAIlQgAAEwwsdI1kggAAMwwJ8Il2hkAAMAjVCIAADBBO8M1kggAAMzQznCJdgYAAPAIlQgAAEzQznCNJAIAADO0M1wiiQAAwAxJhEvMiQAAAB6hEgEAgAnmRLhGJQIAADOGDzYvPPXUU7JYLBo/frxjX3l5ucaOHatWrVqpRYsWGj58uAoKCpzOy8vL05AhQ9S8eXO1bt1akyZNUnV1tXfB1IIkAgCARuizzz7Tq6++qosvvthp/4QJE/TBBx/o3Xff1caNG3X06FHddNNNjuM2m01DhgxRZWWltm7dqrfffltLlizRzJkzfR4jSQQAACYshuH15onS0lKlpKTo9ddfV8uWLR37i4uL9eabb+r555/X//zP/6h3795avHixtm7dqm3btkmSPv74Y33xxRf6f//v/6lXr14aPHiw5syZowULFqiystIn30sNkggAAMz4qJ1RUlLitFVUVLh827Fjx2rIkCFKSkpy2p+VlaWqqiqn/V27dlX79u2VmZkpScrMzFSPHj0UExPjGJOcnKySkhLt27fPwy+idiQRAADUs/j4eEVGRjq29PR007HvvPOOdu3aVeuY/Px8BQcHKyoqyml/TEyM8vPzHWN+mUDUHK855kvcnQEAgAlf3Z1x+PBhWa1Wx/6QkJBaxx8+fFjjxo1TRkaGQkNDPX/jM4RKBAAAZnzUzrBarU6bWRKRlZWlY8eO6dJLL1VQUJCCgoK0ceNGzZs3T0FBQYqJiVFlZaWKioqczisoKFBsbKwkKTY29rS7NWpe14zxFZIIAAAaieuuu0579uxRdna2Y+vTp49SUlIc/71Zs2Zat26d45zc3Fzl5eUpMTFRkpSYmKg9e/bo2LFjjjEZGRmyWq3q3r27T+OlnQEAgIkzvdhURESELrroIqd94eHhatWqlWP/qFGjNHHiREVHR8tqteqBBx5QYmKirrjiCknSwIED1b17d91xxx2aO3eu8vPzNX36dI0dO9a0AuIpkggAAMw0wmdnvPDCCwoICNDw4cNVUVGh5ORkvfLKK47jgYGBWr16te677z4lJiYqPDxcqampmj17ts9jIYkAAMBEY1j2esOGDU6vQ0NDtWDBAi1YsMD0nA4dOmjNmjXev/lvYE4EAADwCJUIAADMNMJ2RmNCEgEAgAv+/iROb9DOAAAAHqESAQCAGcM4tXlzvh8jiQAAwERjuDujMaOdAQAAPEIlAgAAM9yd4RJJBAAAJiz2U5s35/sz2hkAAMAjVCLgkQt7/VfDb/tanboUqdW5FZoztY+2bWojSQoMtOvOP+5Xn8Rjio07qbLSIGXvPFdLFnZT4X9CHdeY+fQOJXQuVlTLSpWeaKbsnedo8SvdncYATcUtaQUa9Ui+3nv9HC16tG1DhwNfoZ3hEpUIeCQ0tFqHDli18Lkepx0LCbXp/AuK9ZfFF+jBkf30xCOXqV37Us18eofTuH/vaqWnZvTRmFsH6MlH+qhN25N65ImdZ+ojAD5zQc+TGnJ7oQ7uIwH2NzV3Z3iz+bMGTSI2bdqkG264QXFxcbJYLFq1atVvnrNhwwZdeumlCgkJUadOnbRkyZJ6jxOny9oWoz+/1lWZP1UffulkWTNNH5+oLevjdCSvhXL3tdTC53uoc7dinRtz0jFu1YrzlbuvpX7Ib66cvdF698+d1OXC4woM9PMmIvxKaHObprz8rV6c1E4nigMbOhz4Ws06Ed5sfqxBk4iysjL17NnT5ZPIfunQoUMaMmSIBgwYoOzsbI0fP1733HOPPvroo3qOFN4KD6+S3S6VnmhW6/EWEZXqP/CIcva0lM1GgQxNR9qTR7RjnVW7N0c0dCjAGdegcyIGDx6swYMH13n8okWLlJCQoOeee06S1K1bN23ZskUvvPCCkpOTaz2noqJCFRUVjtclJSXeBQ23NQu2aeT9OdqY0VY/nnROIkbe/4V+P/wbhYbZlLO3pR57+PIGihJw37VDj6tTjx/1wPWdGzoU1BMWm3KtSf3Jl5mZqaSkJKd9ycnJyszMND0nPT1dkZGRji0+Pr6+w8QvBAbaNW1OlmQxtOCZ0+dP/G3Z+Xrgrn7607grZLdJD83cLb+fiQS/cG5cpe6bfVRPp7VXVUWT+qcU7jB8sPmxJnV3Rn5+vmJiYpz2xcTEqKSkRD/++KPCwsJOO2fatGmaOHGi43VJSQmJxBkSGGjX1MezdG7sj3rkgcTTqhCSVFIcopLiEB093EKHv2mhpf/4l7pedFz790Y3QMRA3XW6+Ee1PLdaCz760rEvMEjqcUWZbhz5H/3+vItlt1saMEKg/jWpJMITISEhCgkJaegwzjo1CURcfJmmpSXqREnwb54TEHAqZW/WjImVaPyyN7fQmAEXOO176IXDOnwgVCsXnEsC4SdoZ7jWpJKI2NhYFRQUOO0rKCiQ1WqttQqB+hMaVq24dmWO17FtTqpj52KdKGmmwv+E6pEnd+r8C4r12KTLFRhgqGV0uSTpREmwqqsD1KX7cXXuVqQv/h2tEyeaqU3bk7pj9H4d/a65cva2bKiPBdTZj2WB+jbX+d+d8pMBOnH89P1owniKp0tNKolITEzUmjVrnPZlZGQoMTGxgSI6e3XuWqSnFvw8F2X0uC8kSf/6sJ2WvdlFV1xzKtl7eekmp/Omjk3Unt3nqLw8UFf2/14p9+QqNNSmwv+GKGtba61Y0lnVVdwmBwBNQYMmEaWlpTpw4IDj9aFDh5Sdna3o6Gi1b99e06ZN05EjR7R06VJJ0r333quXX35ZkydP1t13363169dr5cqV+vDDDxvqI5y19uw+R0OuvMH0uKtjkvTtQaseeeBKX4cFNKjJN3dq6BDgY7QzXGvQJGLnzp0aMGCA43XNBMjU1FQtWbJE33//vfLy8hzHExIS9OGHH2rChAl66aWX1K5dO73xxhumt3cCAOAVlr12qUGTiP79+8tw0S+qbTXK/v37a/fu3fUYFQAAqIsmNScCAIAziXaGayQRAACYsRunNm/O92MkEQAAmGFOhEus1QoAADxCJQIAABMWeTknwmeRNE4kEQAAmGHFSpdoZwAAAI9QiQAAwAS3eLpGEgEAgBnuznCJdgYAAPAIlQgAAExYDEMWLyZHenNuU0ASAQCAGftPmzfn+zHaGQAAwCNUIgAAMEE7wzWSCAAAzHB3hkskEQAAmGHFSpeYEwEAADxCJQIAABOsWOkaSQQAAGZoZ7hEOwMAAHiESgQAACYs9lObN+f7M5IIAADM0M5wiXYGAADwCJUIAADMsNiUS1QiAAAwUbPstTebO9LT03XZZZcpIiJCrVu31rBhw5Sbm+s0pry8XGPHjlWrVq3UokULDR8+XAUFBU5j8vLyNGTIEDVv3lytW7fWpEmTVF1d7fX38WskEQAANBIbN27U2LFjtW3bNmVkZKiqqkoDBw5UWVmZY8yECRP0wQcf6N1339XGjRt19OhR3XTTTY7jNptNQ4YMUWVlpbZu3aq3335bS5Ys0cyZM30er8Uw/HzWx6+UlJQoMjJSSR3GKiggpKHDAepF9Td5DR0CUG+qjSpt0D9UXFwsq9VaL+9R87tiQO9pCgoK9fg61dXl+iQr3eNYf/jhB7Vu3VobN25Uv379VFxcrHPPPVfLly/XzTffLEnav3+/unXrpszMTF1xxRX65z//qd///vc6evSoYmJiJEmLFi3SlClT9MMPPyg4ONjjz/NrVCIAADBjSLJ7sf30Z3pJSYnTVlFRUae3Ly4uliRFR0dLkrKyslRVVaWkpCTHmK5du6p9+/bKzMyUJGVmZqpHjx6OBEKSkpOTVVJSon379nnyLZgiiQAAwISv5kTEx8crMjLSsaWnp//me9vtdo0fP15XXXWVLrroIklSfn6+goODFRUV5TQ2JiZG+fn5jjG/TCBqjtcc8yXuzgAAoJ4dPnzYqZ0REvLb7fSxY8dq79692rJlS32G5hWSCAAAzBjycrGpU/9htVrdmhORlpam1atXa9OmTWrXrp1jf2xsrCorK1VUVORUjSgoKFBsbKxjzI4dO5yuV3P3Rs0YX6GdAQCAmZoVK73Z3Ho7Q2lpaXrvvfe0fv16JSQkOB3v3bu3mjVrpnXr1jn25ebmKi8vT4mJiZKkxMRE7dmzR8eOHXOMycjIkNVqVffu3b34Mk5HJQIAgEZi7NixWr58uf7xj38oIiLCMYchMjJSYWFhioyM1KhRozRx4kRFR0fLarXqgQceUGJioq644gpJ0sCBA9W9e3fdcccdmjt3rvLz8zV9+nSNHTu2Tm0Ud5BEAABgxi7J4uX5bli4cKEkqX///k77Fy9erLvuukuS9MILLyggIEDDhw9XRUWFkpOT9corrzjGBgYGavXq1brvvvuUmJio8PBwpaamavbs2V58kNqRRAAAYMKTVSd/fb476rJ0U2hoqBYsWKAFCxaYjunQoYPWrFnj1nt7gjkRAADAI1QiAAAww6PAXSKJAADADEmES7QzAACAR6hEAABghkqESyQRAACYOcO3eDY1JBEAAJg407d4NjXMiQAAAB6hEgEAgBnmRLhEEgEAgBm7IVm8SATs/p1E0M4AAAAeoRIBAIAZ2hkukUQAAGDKyyRC/p1E0M4AAAAeoRIBAIAZ2hkukUQAAGDGbsirlgR3ZwAAAJyOSgQAAGYM+6nNm/P9GEkEAABmmBPhEkkEAABmmBPhEnMiAACAR6hEAABghnaGSyQRAACYMeRlEuGzSBol2hkAAMAjVCIAADBDO8MlkggAAMzY7ZK8WOvB7t/rRNDOAAAAHqESAQCAGdoZLpFEAABghiTCJdoZAADAI1QiAAAww7LXLpFEAABgwjDsMrx4Eqc35zYFJBEAAJgxDO+qCcyJAAAAOB2VCAAAzBhezonw80oESQQAAGbsdsnixbwGP58TQTsDAAB4hEoEAABmaGe4RBIBAIAJw26X4UU7w99v8aSdAQAAPEIlAgAAM7QzXCKJAADAjN2QLCQRZmhnAAAAj1CJAADAjGFI8madCP+uRJBEAABgwrAbMrxoZxgkEQAAnKUMu7yrRHCLJwAAwGmoRAAAYIJ2hmskEQAAmKGd4dJZl0TUZIXV9soGjgSoP9VGVUOHANSbap36+T4Tf+VXq8qrtaZqYvVXZ10SceLECUnShsOvN3AkAABvnDhxQpGRkfVy7eDgYMXGxmpL/hqvrxUbG6vg4GAfRNX4WAx/b9j8it1u19GjRxURESGLxdLQ4ZwVSkpKFB8fr8OHD8tqtTZ0OIBP8fN95hmGoRMnTiguLk4BAfV3f0B5ebkqK72vWgcHBys0NNQHETU+Z10lIiAgQO3atWvoMM5KVquVf2Tht/j5PrPqqwLxS6GhoX77y99XuMUTAAB4hCQCAAB4hCQC9S4kJESPPvqoQkJCGjoUwOf4+cbZ7KybWAkAAHyDSgQAAPAISQQAAPAISQQAAPAISQQAAPAISQR8YsGCBTrvvPMUGhqqvn37aseOHS7Hv/vuu+ratatCQ0PVo0cPrVnj/dKyQH3YtGmTbrjhBsXFxclisWjVqlW/ec6GDRt06aWXKiQkRJ06ddKSJUvqPU6gIZBEwGsrVqzQxIkT9eijj2rXrl3q2bOnkpOTdezYsVrHb926VbfeeqtGjRql3bt3a9iwYRo2bJj27t17hiMHfltZWZl69uypBQsW1Gn8oUOHNGTIEA0YMEDZ2dkaP3687rnnHn300Uf1HClw5nGLJ7zWt29fXXbZZXr55ZclnXo+SXx8vB544AFNnTr1tPF/+MMfVFZWptWrVzv2XXHFFerVq5cWLVp0xuIG3GWxWPTee+9p2LBhpmOmTJmiDz/80CkpHjFihIqKirR27dozECVw5lCJgFcqKyuVlZWlpKQkx76AgAAlJSUpMzOz1nMyMzOdxktScnKy6XigKeHnG2cTkgh45T//+Y9sNptiYmKc9sfExCg/P7/Wc/Lz890aDzQlZj/fJSUl+vHHHxsoKqB+kEQAAACPkETAK+ecc44CAwNVUFDgtL+goECxsbG1nhMbG+vWeKApMfv5tlqtCgsLa6CogPpBEgGvBAcHq3fv3lq3bp1jn91u17p165SYmFjrOYmJiU7jJSkjI8N0PNCU8PONswlJBLw2ceJEvf7663r77beVk5Oj++67T2VlZRo5cqQk6c4779S0adMc48eNG6e1a9fqueee0/79+zVr1izt3LlTaWlpDfURAFOlpaXKzs5Wdna2pFO3cGZnZysvL0+SNG3aNN15552O8ffee68OHjyoyZMna//+/XrllVe0cuVKTZgwoSHCB+qXAfjA/Pnzjfbt2xvBwcHG5Zdfbmzbts1x7NprrzVSU1Odxq9cudK44IILjODgYOPCCy80PvzwwzMcMVA3n3zyiSHptK3mZzo1NdW49tprTzunV69eRnBwsNGxY0dj8eLFZzxu4ExgnQgAAOAR2hkAAMAjJBEAAMAjJBEAAMAjJBEAAMAjJBEAAMAjJBEAAMAjJBEAAMAjJBEAAMAjJBFAA7jrrrs0bNgwx+v+/ftr/PjxZzyODRs2yGKxqKioyHSMxWLRqlWr6nzNWbNmqVevXl7F9c0338hisTiWmgbQOJFEAD+56667ZLFYZLFYFBwcrE6dOmn27Nmqrq6u9/f++9//rjlz5tRpbF1+8QPAmRDU0AEAjcmgQYO0ePFiVVRUaM2aNRo7dqyaNWvm9ACxGpWVlQoODvbJ+0ZHR/vkOgBwJlGJAH4hJCREsbGx6tChg+677z4lJSXp/fffl/RzC+KJJ55QXFycunTpIkk6fPiwbrnlFkVFRSk6OlpDhw7VN99847imzWbTxIkTFRUVpVatWmny5Mn69SNrft3OqKio0JQpUxQfH6+QkBB16tRJb775pr755hsNGDBAktSyZUtZLBbdddddkk49gj09PV0JCQkKCwtTz5499de//tXpfdasWaMLLrhAYWFhGjBggFOcdTVlyhRdcMEFat68uTp27KgZM2aoqqrqtHGvvvqq4uPj1bx5c91yyy0qLi52Ov7GG2+oW7duCg0NVdeuXfXKK6+4HQuAhkUSAbgQFhamyspKx+t169YpNzdXGRkZWr16taqqqpScnKyIiAht3rxZn376qVq0aKFBgwY5znvuuee0ZMkSvfXWW9qyZYsKCwv13nvvuXzfO++8U3/5y180b9485eTk6NVXX1WLFi0UHx+vv/3tb5Kk3Nxcff/993rppZckSenp6Vq6dKkWLVqkffv2acKECbr99tu1ceNGSaeSnZtuukk33HCDsrOzdc8992jq1KlufycRERFasmSJvvjiC7300kt6/fXX9cILLziNOXDggFauXKkPPvhAa9eu1e7du3X//fc7ji9btkwzZ87UE088oZycHD355JOaMWOG3n77bbfjAdCAGvgpokCjkZqaagwdOtQwDMOw2+1GRkaGERISYjz88MOO4zExMUZFRYXjnD//+c9Gly5dDLvd7thXUVFhhIWFGR999JFhGIbRpk0bY+7cuY7jVVVVRrt27RzvZRinHpc+btw4wzAMIzc315BkZGRk1BpnzaOpjx8/7thXXl5uNG/e3Ni6davT2FGjRhm33nqrYRiGMW3aNKN79+5Ox6dMmXLatX5NkvHee++ZHn/mmWeM3r17O14/+uijRmBgoPHdd9859v3zn/80AgICjO+//94wDMM4//zzjeXLlztdZ86cOUZiYqJhGIZx6NAhQ5Kxe/du0/cF0PCYEwH8wurVq9WiRQtVVVXJbrfrtttu06xZsxzHe/To4TQP4vPPP9eBAwcUERHhdJ3y8nJ9/fXXKi4u1vfff6++ffs6jgUFBalPnz6ntTRqZGdnKzAwUNdee22d4z5w4IBOnjyp3/3ud077Kysrdckll0iScnJynOKQpMTExDq/R40VK1Zo3rx5+vrrr1VaWqrq6mpZrVanMe3bt1fbtm2d3sdutys3N1cRERH6+uuvNWrUKI0ePdoxprq6WpGRkW7HA6DhkEQAvzBgwAAtXLhQwcHBiouLU1CQ8/9FwsPDnV6Xlpaqd+/eWrZs2WnXOvfccz2KISwszO1zSktLJUkffvih0y9v6dQ8D1/JzMxUSkqKHnvsMSUnJysyMlLvvPOOnnvuObdjff31109LagIDA30WK4D6RxIB/EJ4eLg6depU5/GXXnqpVqxYodatW5/213iNNm3aaPv27erXr5+kU39xZ2Vl6dJLL611fI8ePWS327Vx40YlJSWddrymEmKz2Rz7unfvrpCQEOXl5ZlWMLp16+aYJFpj27Ztv/0hf2Hr1q3q0KGD/vSnPzn2ffvtt6eNy8vL09GjRxUXF+d4n4CAAHXp0kUxMTGKi4vTwYMHlZKS4tb7A2hcmFgJeCElJUXnnHOOhg4dqs2bN+vQoUPasGGDHnzwQX333XeSpHHjxumpp57SqlWrtH//ft1///0u13g477zzlJqaqrvvvlurVq1yXHPlypWSpA4dOshisWj16tX64YcfVFpaqoiICD388MOaMGGC3n77bX399dfatWuX5s+f75iseO+99+qrr77SpEmTlJubq+XLl2vJkiVufd7OnTsrLy9P77zzjr7++mvNmzev1kmioaGhSk1N1eeff67NmzfrwQcf1C233KLY2FhJ0mOPPab09HTNmzdPX375pfbs2aPFixfr+eefdyseAA2LJALwQvPmzbVp0ya1b99eN910k7p166ZRo0apvLzcUZl46KGHdMcddyg1NVWJiYmKiIjQ//7v/7q87sKFC3XzzTfr/vvvV9euXTV69GiVlZVJktq2bavHHntMU6dOVUxMjNLS0iRJc+bM0YwZM5Senq5u3bpp0KBB+vDDD5WQkCDp1DyFv/3tb1q1apV69uypRYsW6cknn3Tr8954442aMGGC0tLS1KtXL23dulUzZsw4bVynTp1000036frrr9fAgQN18cUXO93Cec899+iNN97Q4sWL1aNHD1177bVasmSJI1YATYPFMJvdBQAA4AKVCAAA4BGSCAAA4BGSCAAA4BGSCAAA4BGSCAAA4BGSCAAA4BGSCAAA4BGSCAAA4BGSCAAA4BGSCAAA4BGSCAAA4JH/D3GIIzqS9G5LAAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# Select a model from: https://scikit-learn.org/stable/supervised_learning.html#supervised-learning\n",
    "\n",
    "# For example: create a clf variable containing a RandomForestClassifier() or DecisionTreeClassifier(). \n",
    "# If you select another model, you will have to import it first (look above)\n",
    "# Don't hesitate to tweak its parameters as you like ! You can experiment.\n",
    "# Use .fit() on clf variable with your X_train and Y_train to train the model.\n",
    "clf = RandomForestClassifier().fit(X_train, Y_train)\n",
    "\n",
    "print(\"=== QUESTION 2-3 ===\")\n",
    "# Using the .score() method of the clf variable (your trained model), print its accuracy on X_test, Y_test\n",
    "print(clf.score(X_test, Y_test))\n",
    "# Plot the confusion matrix using ConfusionMatrixDisplay.from_estimator() with clf, X_test, Y_test arguments)\n",
    "display = ConfusionMatrixDisplay.from_estimator(clf, X_test, Y_test)\n",
    "plt.show(display.plot())\n"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Tasks: Correct the previous issue\n",
    "\n",
    "You will now try to downsample your majority class (non-diabetes) to the level of minority class (diabetes) to have a 50/50 ratio and re-do a train/test split.  \n",
    "Then you will re-train a new model with the new ratio-corrected data and get accuracy+confusion matrix plot. Don't only predict the class, but also show the prediction probability for all data in the test set !\n",
    "\n",
    "### Questions\n",
    "\n",
    "1. What accuracy-score do you get with the new model and what conclusion can you take.\n",
    "2. What do you observe on the confusion matrix and what conclusion can you take.\n",
    "3. Did you managed to print the probability of each prediction ? What's the shape of the prediction probability output ? Is there a high variance between the different test entries in probability ?\n",
    "\n",
    "![image](https://i.imgur.com/D6iAllf.png)\n",
    "\n",
    "Graphical Recap of what we did so far and what we are going to do ! Because there is a lot of variable and it's easy to be lost.  \n",
    "  \n",
    "![image](https://i.imgur.com/I8FyMhD.png)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=== QUESTION 1-3 ===\n",
      "0.7123655913978495\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "array([[0.26, 0.74],\n",
       "       [0.46, 0.54],\n",
       "       [0.83, 0.17],\n",
       "       [0.54, 0.46],\n",
       "       [0.17, 0.83],\n",
       "       [0.63, 0.37],\n",
       "       [1.  , 0.  ],\n",
       "       [1.  , 0.  ],\n",
       "       [0.8 , 0.2 ],\n",
       "       [0.71, 0.29]])"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# So here because it's a bit complicated I give you directly the function called downsample_majority_class() to do the downsampling.\n",
    "# But you can try to read the code and understand how it works !\n",
    "def downsample_majority_class(X, Y):\n",
    "    np.random.seed(777)\n",
    "    # Get index of rows for each class (0/1)\n",
    "    i_class0 = np.where(Y == 0)[0]\n",
    "    i_class1 = np.where(Y == 1)[0]\n",
    "    # Total Number of observations in each class\n",
    "    n_class0 = len(i_class0)\n",
    "    n_class1 = len(i_class1)\n",
    "    # For every observation of class 1, randomly select (sample) from index list of class 0 without replacement\n",
    "    i_class0_downsampled = np.random.choice(i_class0, size=n_class1, replace=False)\n",
    "    # Join together the downsampled class 0's target vector with class 1's target vector\n",
    "    Y_down = np.hstack((Y[i_class0_downsampled], Y[i_class1]))\n",
    "    X_down = np.vstack((X[i_class0_downsampled], X[i_class1]))\n",
    "    return X_down, Y_down\n",
    "\n",
    "# Call the function downsample_majority_class() on X and Y and store our new dataset in X_down, Y_down\n",
    "X_down, Y_down = downsample_majority_class(X, Y)\n",
    "\n",
    "# Re-do a train/test split (with train_test_split() ) but this time on the new X_down and Y_down data (downsampled).\n",
    "# Call the new variables X_train_down, X_test_down, Y_train_down, Y_test_down for example.\n",
    "X_train_down, X_test_down, Y_train_down, Y_test_down = train_test_split(X_down, Y_down, test_size=0.4, random_state=777)\n",
    "\n",
    "# Recreate a model called clf_down with RandomForestClassifier() or DecisionTreeClassifier() for example.\n",
    "# And .fit() it this time on X_train_down, Y_train_down\n",
    "clf_down = RandomForestClassifier(class_weight='balanced', random_state=777).fit(X_train_down, Y_train_down)\n",
    "\n",
    "print(\"=== QUESTION 1-3 ===\")\n",
    "# Calculate its accuracy using the .score() methods of the clf_down variable with X_test_down, Y_test_down as arguments.\n",
    "print(clf_down.score(X_test_down, Y_test_down))\n",
    "# Plot the confusion matrix as before, but with clf_down, X_test_down, Y_test_down\n",
    "display2 = ConfusionMatrixDisplay.from_estimator(clf_down, X_test_down, Y_test_down)  \n",
    "plt.show(display2.plot())  \n",
    "\n",
    "# Print the prediction probability of the first 10 test data points using clf_down.predict_proba(X_test_down) of the model.\n",
    "clf_down.predict_proba(X_test_down)[:10]\n",
    "\n"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Part 4 - Evaluate your model\n",
    "\n",
    "### Tasks: See all common metrics and evaluate both models !\n",
    "\n",
    "In this last part you will have to calculate all relevant metrics for a binary classification to compare your two models.\n",
    "Make a table containing the results for both models in terms of: accuracy, balance accuracy, F1 Score, sensitivity (recall), specificity, Precision and confusion matrix data (True Pos., True Neg., False Pos., False Neg.)  \n",
    "\n",
    "### Questions:\n",
    "\n",
    "1. Which model have the accuracy ?\n",
    "2. Do you know the difference between accuracy and balanced accuracy ? What model have the best balanced accuracy\n",
    "3. Which model have the best F1-Score and sensitivity ? Do you know how is F1-Score calculated ?\n",
    "4. Eventually, which model is better according to you based on the metrics ?\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Calculate some relevant metrics for a binary classification\n",
    "# Using Scikit score functions: https://scikit-learn.org/stable/modules/model_evaluation.html\n",
    "# Complete \"get_all_metrics\" function to calculate: Accuracy (ac), Balanced Accuracy (bac), recall (re), precision (pr),  F1 Score (f1)\n",
    "\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.metrics import balanced_accuracy_score\n",
    "from sklearn.metrics import recall_score\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn.metrics import precision_score\n",
    "from sklearn.metrics import f1_score\n",
    "\n",
    "def get_all_metrics(clf, X_test, Y_test):\n",
    "\t\"\"\"\"Function that returns all the metrics for a given classifier and test data\"\"\"\n",
    "\t# Use classifier to predict the Y label of X_test data with .predict()\n",
    "\ty_pred = clf.predict(X_test)\n",
    "\n",
    "\t# Confusion Matrix results\n",
    "\ttn, fp, fn, tp = confusion_matrix(Y_test, y_pred).ravel()\n",
    "\n",
    "\t# Calculate the metrics: balanced accuracy, recall, precision, F1-score. Accuracy and Specificity are already done.\n",
    "\tac = accuracy_score(Y_test, y_pred)\n",
    "\tbac = balanced_accuracy_score(Y_test, y_pred)\n",
    "\tre = recall_score(Y_test, y_pred)\n",
    "\tpr = precision_score(Y_test, y_pred)\n",
    "\tf1 = f1_score(Y_test, y_pred)\n",
    "\ttry: sp = tn/float(tn+fp)\n",
    "\texcept: sp = 0\n",
    "\n",
    "\treturn [bac, ac, f1, re, sp, pr, tp, tn, fp, fn]\n",
    "\n",
    "# Here is a function that create a table with the results of the two models, you don't need to touch it, just call it in the next cell.\n",
    "def make_table_clf(clf, clf_down, X_test, Y_test, X_test_down, Y_test_down):\n",
    "\t\"\"\"\"Use the get_all_metrics function to make a table to compare two models\"\"\"\n",
    "\tresults = get_all_metrics(clf, X_test, Y_test)\n",
    "\tresults_down = get_all_metrics(clf_down, X_test_down, Y_test_down)\n",
    "\tdf = pd.DataFrame([[i for i in results], [j for j in results_down]], columns=[\"Balanced-Accuracy\", \"Accuracy\", \"F1-Score\", \"Sensitivity (Recall)\", \"Specificity\", \"Precision\", \"TP\",\" TN\", \"FP\", \"FN\"], index=[\"CLF\",\"CLF DownSampled\"])\n",
    "\treturn df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=== QUESTION 1-4 ===\n"
     ]
    },
    {
     "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>Balanced-Accuracy</th>\n",
       "      <th>Accuracy</th>\n",
       "      <th>F1-Score</th>\n",
       "      <th>Sensitivity (Recall)</th>\n",
       "      <th>Specificity</th>\n",
       "      <th>Precision</th>\n",
       "      <th>TP</th>\n",
       "      <th>TN</th>\n",
       "      <th>FP</th>\n",
       "      <th>FN</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>CLF</th>\n",
       "      <td>0.513796</td>\n",
       "      <td>0.909091</td>\n",
       "      <td>0.058824</td>\n",
       "      <td>0.031496</td>\n",
       "      <td>0.996097</td>\n",
       "      <td>0.444444</td>\n",
       "      <td>4</td>\n",
       "      <td>1276</td>\n",
       "      <td>5</td>\n",
       "      <td>123</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>CLF DownSampled</th>\n",
       "      <td>0.711941</td>\n",
       "      <td>0.712366</td>\n",
       "      <td>0.722078</td>\n",
       "      <td>0.727749</td>\n",
       "      <td>0.696133</td>\n",
       "      <td>0.716495</td>\n",
       "      <td>139</td>\n",
       "      <td>126</td>\n",
       "      <td>55</td>\n",
       "      <td>52</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 Balanced-Accuracy  Accuracy  F1-Score  Sensitivity (Recall)  \\\n",
       "CLF                       0.513796  0.909091  0.058824              0.031496   \n",
       "CLF DownSampled           0.711941  0.712366  0.722078              0.727749   \n",
       "\n",
       "                 Specificity  Precision   TP    TN  FP   FN  \n",
       "CLF                 0.996097   0.444444    4  1276   5  123  \n",
       "CLF DownSampled     0.696133   0.716495  139   126  55   52  "
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print(\"=== QUESTION 1-4 ===\")\n",
    "# Use the make_table_clf function to compare the metrics of both models ! \n",
    "# You need to call it on the clf, clf_down, X_test, Y_test, X_test_down, Y_test_down variables.\n",
    "make_table_clf(clf, clf_down, X_test, Y_test, X_test_down, Y_test_down)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Part 5 - Can we improve the model further ?\n",
    "\n",
    "# Context: \n",
    "**Can we improve the confidence on the performance measure of the model. And can we actually improve its performances ?**\n",
    "* **Cross-Validation**  \n",
    "Instead of a simple Test/Train split, we will do cross-validation. This means that we will train multiple models with different splits, so that all data have been used for training and all for testing. Then we will average the results of all models.\n",
    "For example, if we do a 80/20% train/test split, then we would need to do a 5 fold cross-validation so that each data has been in the test-set at least once.\n",
    "* **Hyper-Parameters Tuning**  \n",
    "Each machine-learning algorithm (for example Random Forest) has a certain number of parameters that can be changed. For example the number of tree used to learn for the data, the maximum depth of the tree, the criterion setting ...  \n",
    "Changing those parameters to find the optimal ones can actually improve the performance of the model !  \n",
    "Here we will learn do hyper-parameters tuning (i.e. searching for the optimal parameters of your model) to improve the performances using what we call a \"Grid Search\"  \n",
    "\n",
    "(Left: cross-validation, Right: hyper-parameters tuning)\n",
    "![image](https://i.imgur.com/ybnTDz1.png)\n",
    "\n",
    "# Task: Fine tune our model to improve its performances\n",
    "* Create a param_grid dictionary variable that will contains all the hyperparameter to test\n",
    "* Call GridSearchCV with our model, the param_grids and the number of cross-val folds to test all combinations\n",
    "* Print what were the best parameters and the final performance score\n",
    "\n",
    "# Questions: \n",
    "1. What is the point of cross-validation ? Does it increase performance ? If not, what is it useful for ? What is the maximum of fold you theoretically do for a cross-validation ?\n",
    "2. For GridSearch, you need to choose what metric you want to maximize, what would you choose ? \n",
    "3. What are the parameters for the try with the best metric ? \n",
    "4. Compare the metric of the best model found using GridSearch to the metrics of the previous clf_down, is it better now ?\n",
    "\n",
    "**Warnings**\n",
    "\n",
    "This can be a pretty long task depending of your computer and the ML algorithm your chose because the program will train and test as many models as parameters combinations multiplied by the number of cross-validation you selected. If it's too long, you can cancel, lower the CV, lower the number of parameters you test or choose a less expensive ML algorithm (for example decision tree, logistic regression...)\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Metrics:  0.7483870967741936\n",
      "Best Parameters:  {'criterion': 'gini', 'max_depth': 5, 'min_samples_leaf': 2, 'min_samples_split': 2, 'n_estimators': 50}\n"
     ]
    },
    {
     "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>mean_fit_time</th>\n",
       "      <th>std_fit_time</th>\n",
       "      <th>mean_score_time</th>\n",
       "      <th>std_score_time</th>\n",
       "      <th>param_criterion</th>\n",
       "      <th>param_max_depth</th>\n",
       "      <th>param_min_samples_leaf</th>\n",
       "      <th>param_min_samples_split</th>\n",
       "      <th>param_n_estimators</th>\n",
       "      <th>params</th>\n",
       "      <th>split0_test_score</th>\n",
       "      <th>split1_test_score</th>\n",
       "      <th>split2_test_score</th>\n",
       "      <th>split3_test_score</th>\n",
       "      <th>split4_test_score</th>\n",
       "      <th>mean_test_score</th>\n",
       "      <th>std_test_score</th>\n",
       "      <th>rank_test_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.102456</td>\n",
       "      <td>0.019348</td>\n",
       "      <td>0.007854</td>\n",
       "      <td>0.000352</td>\n",
       "      <td>gini</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>50</td>\n",
       "      <td>{'criterion': 'gini', 'max_depth': 2, 'min_sam...</td>\n",
       "      <td>0.741935</td>\n",
       "      <td>0.752688</td>\n",
       "      <td>0.763441</td>\n",
       "      <td>0.741935</td>\n",
       "      <td>0.693548</td>\n",
       "      <td>0.738710</td>\n",
       "      <td>0.023947</td>\n",
       "      <td>90</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.186205</td>\n",
       "      <td>0.007127</td>\n",
       "      <td>0.014213</td>\n",
       "      <td>0.000644</td>\n",
       "      <td>gini</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>100</td>\n",
       "      <td>{'criterion': 'gini', 'max_depth': 2, 'min_sam...</td>\n",
       "      <td>0.715054</td>\n",
       "      <td>0.752688</td>\n",
       "      <td>0.790323</td>\n",
       "      <td>0.741935</td>\n",
       "      <td>0.709677</td>\n",
       "      <td>0.741935</td>\n",
       "      <td>0.029052</td>\n",
       "      <td>32</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.343968</td>\n",
       "      <td>0.012393</td>\n",
       "      <td>0.025226</td>\n",
       "      <td>0.000796</td>\n",
       "      <td>gini</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>200</td>\n",
       "      <td>{'criterion': 'gini', 'max_depth': 2, 'min_sam...</td>\n",
       "      <td>0.715054</td>\n",
       "      <td>0.774194</td>\n",
       "      <td>0.784946</td>\n",
       "      <td>0.747312</td>\n",
       "      <td>0.688172</td>\n",
       "      <td>0.741935</td>\n",
       "      <td>0.036146</td>\n",
       "      <td>32</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.097629</td>\n",
       "      <td>0.011176</td>\n",
       "      <td>0.007953</td>\n",
       "      <td>0.000983</td>\n",
       "      <td>gini</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>50</td>\n",
       "      <td>{'criterion': 'gini', 'max_depth': 2, 'min_sam...</td>\n",
       "      <td>0.741935</td>\n",
       "      <td>0.763441</td>\n",
       "      <td>0.741935</td>\n",
       "      <td>0.752688</td>\n",
       "      <td>0.704301</td>\n",
       "      <td>0.740860</td>\n",
       "      <td>0.019943</td>\n",
       "      <td>64</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.185272</td>\n",
       "      <td>0.014604</td>\n",
       "      <td>0.013269</td>\n",
       "      <td>0.000445</td>\n",
       "      <td>gini</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>100</td>\n",
       "      <td>{'criterion': 'gini', 'max_depth': 2, 'min_sam...</td>\n",
       "      <td>0.725806</td>\n",
       "      <td>0.747312</td>\n",
       "      <td>0.779570</td>\n",
       "      <td>0.736559</td>\n",
       "      <td>0.698925</td>\n",
       "      <td>0.737634</td>\n",
       "      <td>0.026426</td>\n",
       "      <td>105</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>238</th>\n",
       "      <td>0.190825</td>\n",
       "      <td>0.001895</td>\n",
       "      <td>0.014505</td>\n",
       "      <td>0.000771</td>\n",
       "      <td>log_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "      <td>5</td>\n",
       "      <td>100</td>\n",
       "      <td>{'criterion': 'log_loss', 'max_depth': 10, 'mi...</td>\n",
       "      <td>0.720430</td>\n",
       "      <td>0.774194</td>\n",
       "      <td>0.784946</td>\n",
       "      <td>0.725806</td>\n",
       "      <td>0.720430</td>\n",
       "      <td>0.745161</td>\n",
       "      <td>0.028368</td>\n",
       "      <td>11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>239</th>\n",
       "      <td>0.370310</td>\n",
       "      <td>0.011401</td>\n",
       "      <td>0.025893</td>\n",
       "      <td>0.000434</td>\n",
       "      <td>log_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "      <td>5</td>\n",
       "      <td>200</td>\n",
       "      <td>{'criterion': 'log_loss', 'max_depth': 10, 'mi...</td>\n",
       "      <td>0.704301</td>\n",
       "      <td>0.752688</td>\n",
       "      <td>0.784946</td>\n",
       "      <td>0.720430</td>\n",
       "      <td>0.709677</td>\n",
       "      <td>0.734409</td>\n",
       "      <td>0.030337</td>\n",
       "      <td>156</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>240</th>\n",
       "      <td>0.091549</td>\n",
       "      <td>0.001264</td>\n",
       "      <td>0.007981</td>\n",
       "      <td>0.000392</td>\n",
       "      <td>log_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "      <td>10</td>\n",
       "      <td>50</td>\n",
       "      <td>{'criterion': 'log_loss', 'max_depth': 10, 'mi...</td>\n",
       "      <td>0.709677</td>\n",
       "      <td>0.763441</td>\n",
       "      <td>0.758065</td>\n",
       "      <td>0.704301</td>\n",
       "      <td>0.698925</td>\n",
       "      <td>0.726882</td>\n",
       "      <td>0.027916</td>\n",
       "      <td>228</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>241</th>\n",
       "      <td>0.182961</td>\n",
       "      <td>0.003084</td>\n",
       "      <td>0.013972</td>\n",
       "      <td>0.000334</td>\n",
       "      <td>log_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "      <td>10</td>\n",
       "      <td>100</td>\n",
       "      <td>{'criterion': 'log_loss', 'max_depth': 10, 'mi...</td>\n",
       "      <td>0.720430</td>\n",
       "      <td>0.747312</td>\n",
       "      <td>0.774194</td>\n",
       "      <td>0.720430</td>\n",
       "      <td>0.693548</td>\n",
       "      <td>0.731183</td>\n",
       "      <td>0.027414</td>\n",
       "      <td>194</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>242</th>\n",
       "      <td>0.376749</td>\n",
       "      <td>0.022903</td>\n",
       "      <td>0.026827</td>\n",
       "      <td>0.000986</td>\n",
       "      <td>log_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "      <td>10</td>\n",
       "      <td>200</td>\n",
       "      <td>{'criterion': 'log_loss', 'max_depth': 10, 'mi...</td>\n",
       "      <td>0.720430</td>\n",
       "      <td>0.779570</td>\n",
       "      <td>0.784946</td>\n",
       "      <td>0.725806</td>\n",
       "      <td>0.709677</td>\n",
       "      <td>0.744086</td>\n",
       "      <td>0.031643</td>\n",
       "      <td>20</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>243 rows × 18 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     mean_fit_time  std_fit_time  mean_score_time  std_score_time  \\\n",
       "0         0.102456      0.019348         0.007854        0.000352   \n",
       "1         0.186205      0.007127         0.014213        0.000644   \n",
       "2         0.343968      0.012393         0.025226        0.000796   \n",
       "3         0.097629      0.011176         0.007953        0.000983   \n",
       "4         0.185272      0.014604         0.013269        0.000445   \n",
       "..             ...           ...              ...             ...   \n",
       "238       0.190825      0.001895         0.014505        0.000771   \n",
       "239       0.370310      0.011401         0.025893        0.000434   \n",
       "240       0.091549      0.001264         0.007981        0.000392   \n",
       "241       0.182961      0.003084         0.013972        0.000334   \n",
       "242       0.376749      0.022903         0.026827        0.000986   \n",
       "\n",
       "    param_criterion param_max_depth param_min_samples_leaf  \\\n",
       "0              gini               2                      1   \n",
       "1              gini               2                      1   \n",
       "2              gini               2                      1   \n",
       "3              gini               2                      1   \n",
       "4              gini               2                      1   \n",
       "..              ...             ...                    ...   \n",
       "238        log_loss              10                      4   \n",
       "239        log_loss              10                      4   \n",
       "240        log_loss              10                      4   \n",
       "241        log_loss              10                      4   \n",
       "242        log_loss              10                      4   \n",
       "\n",
       "    param_min_samples_split param_n_estimators  \\\n",
       "0                         2                 50   \n",
       "1                         2                100   \n",
       "2                         2                200   \n",
       "3                         5                 50   \n",
       "4                         5                100   \n",
       "..                      ...                ...   \n",
       "238                       5                100   \n",
       "239                       5                200   \n",
       "240                      10                 50   \n",
       "241                      10                100   \n",
       "242                      10                200   \n",
       "\n",
       "                                                params  split0_test_score  \\\n",
       "0    {'criterion': 'gini', 'max_depth': 2, 'min_sam...           0.741935   \n",
       "1    {'criterion': 'gini', 'max_depth': 2, 'min_sam...           0.715054   \n",
       "2    {'criterion': 'gini', 'max_depth': 2, 'min_sam...           0.715054   \n",
       "3    {'criterion': 'gini', 'max_depth': 2, 'min_sam...           0.741935   \n",
       "4    {'criterion': 'gini', 'max_depth': 2, 'min_sam...           0.725806   \n",
       "..                                                 ...                ...   \n",
       "238  {'criterion': 'log_loss', 'max_depth': 10, 'mi...           0.720430   \n",
       "239  {'criterion': 'log_loss', 'max_depth': 10, 'mi...           0.704301   \n",
       "240  {'criterion': 'log_loss', 'max_depth': 10, 'mi...           0.709677   \n",
       "241  {'criterion': 'log_loss', 'max_depth': 10, 'mi...           0.720430   \n",
       "242  {'criterion': 'log_loss', 'max_depth': 10, 'mi...           0.720430   \n",
       "\n",
       "     split1_test_score  split2_test_score  split3_test_score  \\\n",
       "0             0.752688           0.763441           0.741935   \n",
       "1             0.752688           0.790323           0.741935   \n",
       "2             0.774194           0.784946           0.747312   \n",
       "3             0.763441           0.741935           0.752688   \n",
       "4             0.747312           0.779570           0.736559   \n",
       "..                 ...                ...                ...   \n",
       "238           0.774194           0.784946           0.725806   \n",
       "239           0.752688           0.784946           0.720430   \n",
       "240           0.763441           0.758065           0.704301   \n",
       "241           0.747312           0.774194           0.720430   \n",
       "242           0.779570           0.784946           0.725806   \n",
       "\n",
       "     split4_test_score  mean_test_score  std_test_score  rank_test_score  \n",
       "0             0.693548         0.738710        0.023947               90  \n",
       "1             0.709677         0.741935        0.029052               32  \n",
       "2             0.688172         0.741935        0.036146               32  \n",
       "3             0.704301         0.740860        0.019943               64  \n",
       "4             0.698925         0.737634        0.026426              105  \n",
       "..                 ...              ...             ...              ...  \n",
       "238           0.720430         0.745161        0.028368               11  \n",
       "239           0.709677         0.734409        0.030337              156  \n",
       "240           0.698925         0.726882        0.027916              228  \n",
       "241           0.693548         0.731183        0.027414              194  \n",
       "242           0.709677         0.744086        0.031643               20  \n",
       "\n",
       "[243 rows x 18 columns]"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.model_selection import GridSearchCV\n",
    "# Create a param_grid dictionary with the parameters you want to test for your model.\n",
    "# For example, if I'm using a KNeighborsClassifier, I would visit https://scikit-learn.org/stable/modules/generated/sklearn.neighbors.KNeighborsClassifier.html#sklearn.neighbors.KNeighborsClassifier\n",
    "# To see all the parameters I can tune, and I would create a dictionary with all the parameters I want to test.\n",
    "# Such as: \n",
    "# param_grid = {\n",
    "# 'n_neighbors': [1, 3, 5, 7, 9],\n",
    "# 'weights': ['uniform', 'distance'],\n",
    "# 'metric': ['euclidean', 'manhattan']\n",
    "# }\n",
    "param_grid = {\n",
    "            'n_estimators': [50, 100, 200],\n",
    "            'max_depth': [2, 5, 10],\n",
    "            'min_samples_leaf': [1, 2, 4],\n",
    "            'min_samples_split': [2, 5, 10],\n",
    "            'criterion': ['gini', 'entropy', \"log_loss\"]\n",
    "            }\n",
    "# Then call GridSearchCV() with your model (KNeighborsClassifier() for example), the param_grid, the number of cross-val folds (cv=NUMBER) and the scoring metric you prefer (scoring=\"your_metric\").\n",
    "# https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.GridSearchCV.html\n",
    "grid_search = GridSearchCV(RandomForestClassifier(), param_grid, cv=5, scoring='balanced_accuracy')\n",
    "# And Call the .fit() method on the grid_search variable with the X_down, Y_down variables as arguments.\n",
    "grid_search.fit(X_down, Y_down)\n",
    "# The code below will print the best parameters and the best score along with the results of all the tests.\n",
    "df_gridsearch = pd.DataFrame(grid_search.cv_results_)\n",
    "print(\"Best Metrics: \", grid_search.best_score_)\n",
    "print(\"Best Parameters: \", grid_search.best_params_)\n",
    "df_gridsearch"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Final Graph of what you did during this practical 🥳\n",
    "![image.png](https://i.imgur.com/QcBKO3Q.png)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Bonus Part: Let's have a small Machine-Learning Contest  \n",
    "\n",
    "### Goal of this part: \n",
    "You can do this part solo or in groups of 2.  \n",
    "This part is a small contest between the different students in the room. You will all train a model on the same data, and the goal is to have the best performing model.\n",
    "\n",
    "### Dataset\n",
    "You will use the [COVID-19 Dataset](https://www.kaggle.com/datasets/meirnizri/covid19-dataset) to predict the death risk of some patients. You can see the meaning of all columns on the link.  \n",
    "You will have 10.000 patients in the training set and 1.000 patients in the test set with unknown label that you will need to predict.  \n",
    "I already processed and encoded all the data for you so you can directly train a ML model on it. I give you the code to import all data.\n",
    "You will have a training dataset with the features and the labels. And you will have a test dataset in which I removed the labels.  \n",
    "\n",
    "### Your task.\n",
    "\n",
    "1. Your task will be to train the best machine-learning model possible using the previous knowledge from this pratical.  \n",
    "2. Then you will predict the labels of the test set (from which I removed the true labels !)  \n",
    "3. Then you will upload the results to the leaderboard (the code is already provided).  \n",
    "\n",
    "You will be able to see how well your model performed (F1-Score) on the mystery dataset (test) and the score of the other students in the room.  \n",
    "Your goal is to be #1 on the leaderboard, i.e. having the best performing model (F1-Score is the main metric)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Training Set with X_train (features) and Y_train (labels)\n",
    "covid_df_train = pd.read_csv(\"https://lbgi.fr/~meyer/TD_ML2/covid_train.csv\")\n",
    "X_covid_train = covid_df_train.drop('DEATH', axis=1).to_numpy()\n",
    "Y_covid_train = covid_df_train['DEATH'].to_numpy()\n",
    "# Testing Set only X_test (features) is provided (true labels are removed).\n",
    "X_covid_test = pd.read_csv(\"https://lbgi.fr/~meyer/TD_ML2/covid_test.csv\").to_numpy()\n",
    "\n",
    "# I converted the dataframe to numpy array because some function requier numpy array."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Now it's your turn ! Train the best model you can on X_covid_train Y_covid_train. And predict the X_covid_test data !  \n",
    "# Tip to optimize your model: try different algorithms and also the gridsearch like in the part 5.\n",
    "# Then submit your data with the next code block !\n",
    "\n",
    "# I will first downsample my majority class.\n",
    "X_covid_down, Y_covid_down = downsample_majority_class(X_covid_train, Y_covid_train)\n",
    "\n",
    "# Create and train your model on the X_covid_down, Y_covid_down variables. Here for example I will use a RandomForestClassifier()\n",
    "clf_covid = RandomForestClassifier(class_weight='balanced', random_state=777).fit(X_covid_down, Y_covid_down)\n",
    "# Predict your X_covid_test\n",
    "Y_pred = clf_covid.predict(X_covid_test)\n",
    "# Now that you have you Y_pred you can go to the next cell to send your prediction\n",
    "# You will get your score and your rank in the leaderboard."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=== LEADERBOARD ===\n",
      "Rank  1  -  Corentin  - F1-Score:\t 0.723152709359606\n"
     ]
    }
   ],
   "source": [
    "# You don't need to touch this code it's function I created for you to sumbit your score to the leaderboard.\n",
    "# And to print the leaderboard. Go at the end of this block.\n",
    "import requests\n",
    "def send_my_score(Y_pred, student_name: str):\n",
    "    Y_pred = Y_pred.tolist()\n",
    "    url = \"https://td_api-1-z7311418.deta.app/score\"\n",
    "    headers = {\"accept\": \"application/json\", \"Content-Type\": \"application/json\"}\n",
    "    querystring = {\"student_name\": student_name}\n",
    "    response = requests.request(\n",
    "        \"POST\", url, params=querystring, json=Y_pred, headers=headers\n",
    "    )\n",
    "    print(response.json()[\"msg\"])\n",
    "    print(response.json()[\"rank_msg\"])\n",
    "    print(\"Your score is: \", response.json()[\"F1\"])\n",
    "    print(\"=== LEADERBOARD ===\")\n",
    "    counter = 1\n",
    "    for key, value in response.json()[\"leaderboard\"].items():\n",
    "        print(\"Rank \", counter, \" - \", key, \" - F1-Score:\\t: \", value)\n",
    "        counter += 1\n",
    "\n",
    "def show_leaderboard():\n",
    "    url = \"https://td_api-1-z7311418.deta.app/\"\n",
    "    headers = {\"accept\": \"application/json\", \"Content-Type\": \"application/json\"}\n",
    "    response = requests.request(\"GET\", url, headers=headers)\n",
    "    print(\"=== LEADERBOARD ===\")\n",
    "    counter = 1\n",
    "    for key, value in response.json()[\"leaderboard\"].items():\n",
    "        print(\"Rank \", counter, \" - \", key, \" - F1-Score:\\t\", value)\n",
    "        counter += 1\n",
    "\n",
    "\n",
    "# All you have to do is to execute show_leaderboard() to \n",
    "# print the current leaderboard. \n",
    "show_leaderboard()\n",
    "\n",
    "# Once you ran the command: Y_pred = MY_MODEL.predict(X_covid_test)\n",
    "# You can execute the following commands to submit your score to the leaderboard. don't forget to change your name !\n",
    "# TIP: most of the time if you have an error, that mean your Y_pred doesn't containt 1000 predictions as expected !\n",
    "#send_my_score(Y_pred, \"Corentin\")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Further Resources\n",
    "## General Python\n",
    "* Probably the best website to learn about some Python lib https://calmcode.io/\n",
    "* Made by the somebody that I like a lot (Vincent D. Warmerdam) and his blog https://koaning.io/\n",
    "* **Self-promo: I also have a website: https://cmeyer.fr/**\n",
    "* Visualisation in Python https://www.python-graph-gallery.com/\n",
    "\n",
    "# AI & Data Specific\n",
    "* Google's Class lists and ressources on ML https://developers.google.com/machine-learning/crash-course\n",
    "* Wanna learn AI ? Check this Roadmap https://i.am.ai/roadmap\n",
    "* Public Dataset to create projects https://github.com/awesomedata/awesome-public-datasets\n",
    "* Machine Learning Competitions with Prize https://www.kaggle.com/"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "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": "3.8.16"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "618c32a8dee67f46364144f4ba38dfe00bd7a891c8a0c80cda06fb6d8e41a78b"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}