Roosevelt Forest - First intuition in a stacked Ensemble model.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# Roosevelt National Forest of northern Colorado Competition\n",
    "\n",
    "In this experiment we will apply some classifiers based on trees and then stack them all together to find out a good classifier\n",
    "\n",
    "Ensemble Methods: RF, AdaBoost"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import os\n",
    "import math\n",
    "\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Global parameters\n",
    "#root folder\n",
    "root_folder='..\\..'\n",
    "#data_folder='.'\n",
    "data_folder_name='Datasets/RooseveltForest'\n",
    "\n",
    "# Variable for data directory\n",
    "DATA_PATH = os.path.abspath(os.path.join(root_folder, data_folder_name))\n",
    "\n",
    "# Both train and test set are in the root data directory\n",
    "train_path = DATA_PATH\n",
    "test_path = DATA_PATH\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Read training and test data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Reading the csv file with the whole dataset\n",
    "data = pd.read_csv(train_path+'\\\\train.csv', sep=',', header=0, index_col=0)\n",
    "# Train data, erasing label column\n",
    "train_data = data.drop('Cover_Type', axis=1)\n",
    "train_label = data[['Cover_Type']]\n",
    "# Read the test data\n",
    "test_data = pd.read_csv(test_path+'\\\\test.csv', sep=',', header=0, index_col=0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data preprocessing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Global variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define some groups of columns based on  type\n",
    "var_numeric = ['Elevation','Aspect','Slope','Horizontal_Distance_To_Hydrology','Vertical_Distance_To_Hydrology',\n",
    "                'Horizontal_Distance_To_Roadways','Hillshade_9am','Hillshade_Noon','Hillshade_3pm',\n",
    "                'Horizontal_Distance_To_Fire_Points']\n",
    "# There are no categoric variables to transfomr\n",
    "var_categoric = []\n",
    "\n",
    "# Numerical columns with missing values\n",
    "# In our first approach, no missing values to consider\n",
    "columns_nan=[]\n",
    "\n",
    "#Defining some relevant group of columns\n",
    "# Columns to be ingored for classification\n",
    "# Soil Type 7,8, 15 y 25\n",
    "columns_to_ignore = ['Soil_Type7','Soil_Type8','Soil_Type15','Soil_Type25']\n",
    "\n",
    "num_columns_nan=[i for i in columns_nan if i not in columns_to_ignore]\n",
    "enconde_cols_name=[i for i in var_categoric if i not in columns_to_ignore]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Feature engineering and selection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Horizontal to roadways and to Fire poins are very simila recpect to arget variable so lets drop horizontal to roadways\n",
    "# Horizontal and vertical to hidrology are correlated, lets drop vertical\n",
    "# Drop soil types where the number of rows are very low: 9, 21, 27, 28, 36\n",
    "# Outliers: \n",
    "columns_unselect=['Soil_Type9','Soil_Type21','Soil_Type27','Soil_Type28','Soil_Type36']\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Manipulate and transform columns in dataframe to be used in the next stages\n",
    "\n",
    "# Delete non relevant columns\n",
    "train_data.drop(columns_to_ignore+columns_unselect, axis=1,inplace=True)\n",
    "test_data.drop(columns_to_ignore+columns_unselect, axis=1,inplace=True)\n",
    "\n",
    "#Filling na values in columns\n",
    "#Not neccesary at this moment\n",
    "\n",
    "#Dealing with ourliers or data errors\n",
    "#Not neccesary at this moment\n",
    "\n",
    "# Define columns to fill nan values and categorial columns to one-hot-encode \n",
    "columns_nan=[train_df.columns.tolist().index(i) for i in num_columns_nan]\n",
    "columns_enconde=[train_df.columns.tolist().index(i) for i in enconde_cols_name]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train data:  (15120, 45)\n",
      "Train labels:  (15120, 1)\n",
      "Test data:  (565892, 45)\n"
     ]
    }
   ],
   "source": [
    "print('Train data: ',train_data.shape)\n",
    "print('Train labels: ',train_label.shape)\n",
    "print('Test data: ',test_data.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preparing the data for the algorithm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X:  (15120, 45)\n",
      "y:  (15120,)\n",
      "test:  (565892, 45)\n"
     ]
    }
   ],
   "source": [
    "# Tranform to numpy array\n",
    "X = train_data.values.astype('float64')\n",
    "y = train_label.values.ravel()\n",
    "test_ds= test_data.values.astype('float64')\n",
    "\n",
    "#Convertimos todos los valores a float incluida la columna boolean\n",
    "# To avoid error in the imputer action if it was neccesary\n",
    "#X= np.vstack(X[:, :]).astype(np.float)\n",
    "\n",
    "print('X: ', X.shape)\n",
    "print('y: ', y.shape)\n",
    "print('test: ', test_ds.shape)\n",
    "#X Shape 500,000 x 8"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[2.596e+03, 5.100e+01, 3.000e+00, ..., 0.000e+00, 0.000e+00,\n",
       "        0.000e+00],\n",
       "       [2.590e+03, 5.600e+01, 2.000e+00, ..., 0.000e+00, 0.000e+00,\n",
       "        0.000e+00],\n",
       "       [2.804e+03, 1.390e+02, 9.000e+00, ..., 0.000e+00, 0.000e+00,\n",
       "        0.000e+00],\n",
       "       ...,\n",
       "       [2.492e+03, 1.340e+02, 2.500e+01, ..., 0.000e+00, 0.000e+00,\n",
       "        0.000e+00],\n",
       "       [2.487e+03, 1.670e+02, 2.800e+01, ..., 0.000e+00, 0.000e+00,\n",
       "        0.000e+00],\n",
       "       [2.475e+03, 1.970e+02, 3.400e+01, ..., 0.000e+00, 0.000e+00,\n",
       "        0.000e+00]])"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Building and training an algorithm"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Creating the train and test datasets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\edumu\\Anaconda3\\lib\\site-packages\\sklearn\\utils\\validation.py:595: DataConversionWarning: Data with input dtype int64 was converted to float64 by StandardScaler.\n",
      "  warnings.warn(msg, DataConversionWarning)\n"
     ]
    }
   ],
   "source": [
    "# Splitting the dataset into the Training set and Test set\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.4, random_state = 0)\n",
    "\n",
    "# Feature Scaling\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "sc = StandardScaler()\n",
    "X_train = sc.fit_transform(X_train)\n",
    "X_test = sc.transform(X_test)\n",
    "test_ds= sc.transform(test_ds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X Train Shape: (9072, 45)\n",
      "X Test Shape: (6048, 45)\n",
      "Test Shape: (565892, 45)\n"
     ]
    }
   ],
   "source": [
    "print('X Train Shape:', X_train.shape)\n",
    "print('X Test Shape:', X_test.shape)\n",
    "print('Test Shape:', test_ds.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<bound method DataFrame.info of Empty DataFrame\n",
       "Columns: [RF]\n",
       "Index: []>"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train_l2 = pd.DataFrame(columns=['RF'])\n",
    "X_train_l2.info"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train the algorithm: Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 3 candidates, totalling 15 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  15 out of  15 | elapsed:   11.2s finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=5, error_score='raise-deprecating',\n",
       "       estimator=RandomForestClassifier(bootstrap=True, class_weight=None, criterion='entropy',\n",
       "            max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
       "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "            min_samples_leaf=1, min_samples_split=2,\n",
       "            min_weight_fraction_leaf=0.0, n_estimators='warn', n_jobs=None,\n",
       "            oob_score=False, random_state=0, verbose=0, warm_start=False),\n",
       "       fit_params=None, iid='warn', n_jobs=-1,\n",
       "       param_grid=[{'n_estimators': [50, 100, 125]}],\n",
       "       pre_dispatch='2*n_jobs', refit=True, return_train_score=True,\n",
       "       scoring='accuracy', verbose=1)"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Fitting Decision Tree to the Training set\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "\n",
    "parameter_candidates = [{'n_estimators': [50, 100, 125]}]\n",
    "\n",
    "classifier = GridSearchCV(estimator=RandomForestClassifier(criterion = 'entropy', max_depth=None,min_samples_split=2,\n",
    "                                                           min_samples_leaf=1, max_leaf_nodes=None,random_state = 0),\n",
    "                          param_grid=parameter_candidates, scoring='accuracy', n_jobs=-1, cv=5, verbose=1, refit=True, \n",
    "                          return_train_score=True)\n",
    "\n",
    "# Train the classifier on data1's feature and target data\n",
    "classifier.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'n_estimators': 125}\n",
      "RandomForestClassifier(bootstrap=True, class_weight=None, criterion='entropy',\n",
      "            max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
      "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
      "            min_samples_leaf=1, min_samples_split=2,\n",
      "            min_weight_fraction_leaf=0.0, n_estimators=125, n_jobs=None,\n",
      "            oob_score=False, random_state=0, verbose=0, warm_start=False)\n"
     ]
    }
   ],
   "source": [
    "#Show the results of the hyperparameter tunning\n",
    "print(classifier.best_params_)\n",
    "print(classifier.best_estimator_)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import StratifiedKFold\n",
    "\n",
    "y_preds, y_actual=[],[]\n",
    "for train_index, test_index in StratifiedKFold(n_splits=10, random_state=1234, shuffle=True).split(X, y):\n",
    "    y_preds.append(classifier.best_estimator_.predict_proba(X[test_index]))\n",
    "    y_actual.append(y[test_index])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(6048,)\n",
      "(565892,)\n",
      "0.8466\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>RF</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6018</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6019</th>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6020</th>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6021</th>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6022</th>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6023</th>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6024</th>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6025</th>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6026</th>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6027</th>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6028</th>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6029</th>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6030</th>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6031</th>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6032</th>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6033</th>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6034</th>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6035</th>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6036</th>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6037</th>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6038</th>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6039</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6040</th>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6041</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6042</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6043</th>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6044</th>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6045</th>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6046</th>\n",
       "      <td>7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6047</th>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>6048 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      RF\n",
       "0      4\n",
       "1      4\n",
       "2      4\n",
       "3      4\n",
       "4      4\n",
       "5      1\n",
       "6      4\n",
       "7      7\n",
       "8      1\n",
       "9      7\n",
       "10     7\n",
       "11     3\n",
       "12     1\n",
       "13     4\n",
       "14     2\n",
       "15     3\n",
       "16     7\n",
       "17     3\n",
       "18     4\n",
       "19     3\n",
       "20     4\n",
       "21     7\n",
       "22     3\n",
       "23     5\n",
       "24     7\n",
       "25     5\n",
       "26     2\n",
       "27     5\n",
       "28     3\n",
       "29     1\n",
       "...   ..\n",
       "6018   1\n",
       "6019   7\n",
       "6020   4\n",
       "6021   4\n",
       "6022   6\n",
       "6023   3\n",
       "6024   7\n",
       "6025   2\n",
       "6026   7\n",
       "6027   4\n",
       "6028   5\n",
       "6029   5\n",
       "6030   6\n",
       "6031   7\n",
       "6032   2\n",
       "6033   6\n",
       "6034   2\n",
       "6035   5\n",
       "6036   2\n",
       "6037   3\n",
       "6038   6\n",
       "6039   1\n",
       "6040   6\n",
       "6041   1\n",
       "6042   1\n",
       "6043   2\n",
       "6044   2\n",
       "6045   4\n",
       "6046   7\n",
       "6047   1\n",
       "\n",
       "[6048 rows x 1 columns]"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.metrics import accuracy_score\n",
    "\n",
    "predictions=classifier.best_estimator_.predict(X_test)\n",
    "print(predictions.shape)\n",
    "predictions_test = classifier.best_estimator_.predict(test_ds)\n",
    "print(predictions_test.shape)\n",
    "print(round(accuracy_score(y_test, predictions),4))\n",
    "\n",
    "X_train_l2['RF'] = predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train the algorithm: Decision Tree"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(6048,)\n",
      "(565892,)\n",
      "0.7235\n"
     ]
    }
   ],
   "source": [
    "# Fitting Decision Tree to the Training set\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "\n",
    "classifier = DecisionTreeClassifier(criterion = 'entropy',max_depth=None,min_samples_split=2, min_samples_leaf=1,\n",
    "                                    max_leaf_nodes=None,max_features='auto', random_state = 0)\n",
    "\n",
    "classifier.fit(X_train, y_train)\n",
    "#Show the results of the hyperparameter tunning\n",
    "predictions=classifier.predict(X_test)\n",
    "print(predictions.shape)\n",
    "predictions_test = classifier.predict(test_ds)\n",
    "print(predictions_test.shape)\n",
    "print(round(accuracy_score(y_test, predictions),4))\n",
    "\n",
    "X_train_l2['DT'] = predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train the algorithm: AdaBoosted Decision Tree"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 9 candidates, totalling 45 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  45 out of  45 | elapsed:  4.3min finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=5, error_score='raise-deprecating',\n",
       "       estimator=AdaBoostClassifier(algorithm='SAMME.R',\n",
       "          base_estimator=DecisionTreeClassifier(class_weight=None, criterion='entropy', max_depth=None,\n",
       "            max_features=None, max_leaf_nodes=None,\n",
       "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "            min_samples_leaf=2, min_samples_split=2,\n",
       "            min_weight_fraction_leaf=0.0, presort=False, random_state=0,\n",
       "            splitter='best'),\n",
       "          learning_rate=1.0, n_estimators=50, random_state=0),\n",
       "       fit_params=None, iid='warn', n_jobs=-1,\n",
       "       param_grid=[{'n_estimators': [50, 100, 125], 'learning_rate': [0.25, 0.5, 1.0]}],\n",
       "       pre_dispatch='2*n_jobs', refit=True, return_train_score=True,\n",
       "       scoring='accuracy', verbose=1)"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.ensemble import AdaBoostClassifier\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "\n",
    "# Define the base estimator for the Adaboost algorithm\n",
    "base_classifier = DecisionTreeClassifier(criterion ='entropy',max_depth=None, min_samples_split=2,min_samples_leaf=2,\n",
    "                                                       max_leaf_nodes=None,random_state = 0)\n",
    "\n",
    "parameter_candidates = [{'n_estimators': [50, 100, 125], 'learning_rate': [0.25, 0.5, 1.]}]\n",
    "\n",
    "classifier = GridSearchCV(estimator=AdaBoostClassifier(base_estimator=base_classifier,random_state=0),\n",
    "                          param_grid=parameter_candidates, scoring='accuracy', n_jobs=-1, cv=5, verbose=1, refit=True, \n",
    "                          return_train_score=True)\n",
    "\n",
    "classifier.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'learning_rate': 0.25, 'n_estimators': 100}\n",
      "AdaBoostClassifier(algorithm='SAMME.R',\n",
      "          base_estimator=DecisionTreeClassifier(class_weight=None, criterion='entropy', max_depth=None,\n",
      "            max_features=None, max_leaf_nodes=None,\n",
      "            min_impurity_decrease=0.0, min_impurity_split=None,\n",
      "            min_samples_leaf=2, min_samples_split=2,\n",
      "            min_weight_fraction_leaf=0.0, presort=False, random_state=0,\n",
      "            splitter='best'),\n",
      "          learning_rate=0.25, n_estimators=100, random_state=0)\n",
      "(6048,)\n",
      "(565892,)\n",
      "0.8522\n"
     ]
    }
   ],
   "source": [
    "#Show the results of the hyperparameter tunning\n",
    "print(classifier.best_params_)\n",
    "print(classifier.best_estimator_)\n",
    "\n",
    "predictions=classifier.best_estimator_.predict(X_test)\n",
    "print(predictions.shape)\n",
    "predictions_test = classifier.best_estimator_.predict(test_ds)\n",
    "print(predictions_test.shape)\n",
    "print(round(accuracy_score(y_test, predictions),4))\n",
    "\n",
    "X_train_l2['AB'] = predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train the algorithm: Gradient Boosting Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 27 candidates, totalling 135 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:  2.2min\n",
      "[Parallel(n_jobs=-1)]: Done 135 out of 135 | elapsed: 20.4min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "      Iter       Train Loss   Remaining Time \n",
      "         1       10620.1119            1.40m\n",
      "         2        7733.7413            1.44m\n",
      "         3        5869.8905            1.46m\n",
      "         4        4594.7696            1.47m\n",
      "         5        3625.5944            1.47m\n",
      "         6        2919.8230            1.46m\n",
      "         7        2391.3320            1.45m\n",
      "         8        1977.9951            1.45m\n",
      "         9        1640.1807            1.44m\n",
      "        10        1374.9496            1.42m\n",
      "        20         359.2368            1.29m\n",
      "        30         141.6159            1.09m\n",
      "        40          57.6664           54.84s\n",
      "        50          24.8450           45.95s\n",
      "        60          11.6845           37.46s\n",
      "        70           6.2602           29.63s\n",
      "        80           4.1854           22.43s\n",
      "        90           3.4764           15.99s\n",
      "       100           3.0524           10.53s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=5, error_score='raise-deprecating',\n",
       "       estimator=GradientBoostingClassifier(criterion='friedman_mse', init=None,\n",
       "              learning_rate=0.1, loss='deviance', max_depth=3,\n",
       "              max_features=None, max_leaf_nodes=None,\n",
       "              min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "              min_samples_leaf=1, min_sampl...      subsample=1.0, tol=0.0001, validation_fraction=0.1,\n",
       "              verbose=1, warm_start=False),\n",
       "       fit_params=None, iid='warn', n_jobs=-1,\n",
       "       param_grid=[{'n_estimators': [50, 100, 125], 'learning_rate': [0.5, 0.2, 0.1], 'max_depth': [5, 10, None]}],\n",
       "       pre_dispatch='2*n_jobs', refit=True, return_train_score=True,\n",
       "       scoring=None, verbose=1)"
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Fitting Decision Tree to the Training set\n",
    "from sklearn.ensemble import GradientBoostingClassifier\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "#from sklearn.tree import DecisionTreeClassifier\n",
    "#classifier = GradientBoostingClassifier(n_estimators=100,learning_rate=0.5,max_depth=15,\n",
    "#              min_samples_leaf=1, min_samples_split=2,random_state=0,verbose=1)\n",
    "#Hyperparameter tuning with Grid Search\n",
    "parameter_candidates = [{'n_estimators': [50, 100, 125], 'learning_rate': [0.5,0.2,0.1], 'max_depth':[5,10,None]}]\n",
    "\n",
    "classifier = GridSearchCV(estimator=GradientBoostingClassifier(min_samples_leaf=1, min_samples_split=2,random_state=0,verbose=1),\n",
    "                          param_grid=parameter_candidates, n_jobs=-1, cv=5, verbose=1, return_train_score=True)\n",
    "\n",
    "# Train the classifier on data1's feature and target data\n",
    "classifier.fit(X_train, y_train)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'learning_rate': 0.2, 'max_depth': 10, 'n_estimators': 125}\n",
      "GradientBoostingClassifier(criterion='friedman_mse', init=None,\n",
      "              learning_rate=0.2, loss='deviance', max_depth=10,\n",
      "              max_features=None, max_leaf_nodes=None,\n",
      "              min_impurity_decrease=0.0, min_impurity_split=None,\n",
      "              min_samples_leaf=1, min_samples_split=2,\n",
      "              min_weight_fraction_leaf=0.0, n_estimators=125,\n",
      "              n_iter_no_change=None, presort='auto', random_state=0,\n",
      "              subsample=1.0, tol=0.0001, validation_fraction=0.1,\n",
      "              verbose=1, warm_start=False)\n",
      "(6048,)\n",
      "(565892,)\n",
      "0.8476\n"
     ]
    }
   ],
   "source": [
    "#Show the results of the hyperparameter tunning\n",
    "print(classifier.best_params_)\n",
    "print(classifier.best_estimator_)\n",
    "\n",
    "predictions=classifier.best_estimator_.predict(X_test)\n",
    "print(predictions.shape)\n",
    "predictions_test = classifier.best_estimator_.predict(test_ds)\n",
    "print(predictions_test.shape)\n",
    "print(round(accuracy_score(y_test, predictions),4))\n",
    "\n",
    "X_train_l2['GB'] = predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean:  0.8479719984497024\n",
      "Std:  0.00590787992349919\n"
     ]
    }
   ],
   "source": [
    "# Scoring the model by Cross Validation\n",
    "from sklearn.model_selection import cross_val_score\n",
    "accuracies = cross_val_score(estimator = classifier, X = X_train, y = y_train, cv = 10)\n",
    "mean_score = accuracies.mean()\n",
    "std_score = accuracies.std()\n",
    "print('Mean: ',mean_score)\n",
    "print('Std: ',std_score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 3 candidates, totalling 15 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  15 out of  15 | elapsed:    3.8s finished\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "GridSearchCV(cv=5, error_score='raise-deprecating',\n",
       "       estimator=ExtraTreesClassifier(bootstrap=False, class_weight=None, criterion='entropy',\n",
       "           max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
       "           min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "           min_samples_leaf=1, min_samples_split=2,\n",
       "           min_weight_fraction_leaf=0.0, n_estimators='warn', n_jobs=None,\n",
       "           oob_score=False, random_state=0, verbose=0, warm_start=False),\n",
       "       fit_params=None, iid='warn', n_jobs=-1,\n",
       "       param_grid=[{'n_estimators': [50, 100, 125]}],\n",
       "       pre_dispatch='2*n_jobs', refit=True, return_train_score=True,\n",
       "       scoring='accuracy', verbose=1)"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Fitting Decision Tree to the Training set\n",
    "from sklearn.ensemble import ExtraTreesClassifier\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "\n",
    "parameter_candidates = [{'n_estimators': [50, 100, 125]}]\n",
    "\n",
    "classifier = GridSearchCV(estimator=ExtraTreesClassifier(criterion = 'entropy', max_depth=None,min_samples_split=2,\n",
    "                                                           min_samples_leaf=1, max_leaf_nodes=None,random_state = 0),\n",
    "                          param_grid=parameter_candidates, scoring='accuracy', n_jobs=-1, cv=5, verbose=1, refit=True, \n",
    "                          return_train_score=True)\n",
    "\n",
    "# Train the classifier on data1's feature and target data\n",
    "classifier.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'n_estimators': 125}\n",
      "ExtraTreesClassifier(bootstrap=False, class_weight=None, criterion='entropy',\n",
      "           max_depth=None, max_features='auto', max_leaf_nodes=None,\n",
      "           min_impurity_decrease=0.0, min_impurity_split=None,\n",
      "           min_samples_leaf=1, min_samples_split=2,\n",
      "           min_weight_fraction_leaf=0.0, n_estimators=125, n_jobs=None,\n",
      "           oob_score=False, random_state=0, verbose=0, warm_start=False)\n",
      "(6048,)\n",
      "(565892,)\n",
      "0.848\n"
     ]
    }
   ],
   "source": [
    "#Show the results of the hyperparameter tunning\n",
    "print(classifier.best_params_)\n",
    "print(classifier.best_estimator_)\n",
    "\n",
    "predictions=classifier.best_estimator_.predict(X_test)\n",
    "print(predictions.shape)\n",
    "predictions_test = classifier.best_estimator_.predict(test_ds)\n",
    "print(predictions_test.shape)\n",
    "print(round(accuracy_score(y_test, predictions),4))\n",
    "\n",
    "X_train_l2['ET'] = predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "ename": "ModuleNotFoundError",
     "evalue": "No module named 'sklearn.experimental'",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-67-c05d3f65657c>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m      1\u001b[0m \u001b[1;31m# Fitting Decision Tree to the Training set\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 2\u001b[1;33m \u001b[1;32mfrom\u001b[0m \u001b[0msklearn\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mexperimental\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0menable_hist_gradient_boosting\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      3\u001b[0m \u001b[1;32mfrom\u001b[0m \u001b[0msklearn\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mensemble\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mHistGradientBoostingClassifier\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      4\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      5\u001b[0m \u001b[0mparameter_candidates\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;33m[\u001b[0m\u001b[1;33m{\u001b[0m\u001b[1;34m'max_iter'\u001b[0m\u001b[1;33m:\u001b[0m \u001b[1;33m[\u001b[0m\u001b[1;36m50\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;36m100\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;36m125\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;34m'learning_rate'\u001b[0m\u001b[1;33m:\u001b[0m \u001b[1;33m[\u001b[0m\u001b[1;36m0.5\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m0.2\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;36m0.1\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m}\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mModuleNotFoundError\u001b[0m: No module named 'sklearn.experimental'"
     ]
    }
   ],
   "source": [
    "# Fitting Decision Tree to the Training set\n",
    "from sklearn.experimental import enable_hist_gradient_boosting\n",
    "from sklearn.ensemble import HistGradientBoostingClassifier\n",
    "\n",
    "parameter_candidates = [{'max_iter': [50, 100, 125], 'learning_rate': [0.5,0.2,0.1]}]\n",
    "\n",
    "classifier = GridSearchCV(estimator=GradientBoostingClassifier(min_samples_leaf=20, random_state=0,verbose=1),\n",
    "                          param_grid=parameter_candidates, n_jobs=-1, cv=5, verbose=1, return_train_score=True)\n",
    "\n",
    "# Train the classifier on data1's feature and target data\n",
    "classifier.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Scoring and evaluating the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion Matrix:  [[415  91   1   0   7   1  30]\n",
      " [ 91 368  15   0  38   7   6]\n",
      " [  0   1 457  25  10  58   0]\n",
      " [  0   0   9 522   0  13   0]\n",
      " [  4  27   5   0 485   4   0]\n",
      " [  0   6  56   7   9 481   0]\n",
      " [ 17   2   0   0   0   0 512]]\n",
      "Accuracy:  0.8571428571428571\n"
     ]
    }
   ],
   "source": [
    "# Predicting the Test set results\n",
    "y_pred = classifier.predict(X_test)\n",
    "\n",
    "# Making the Confusion Matrix\n",
    "from sklearn.metrics import confusion_matrix,accuracy_score,roc_auc_score,roc_curve,auc\n",
    "cm = confusion_matrix(y_test, y_pred)\n",
    "print('Confusion Matrix: ',cm)\n",
    "\n",
    "acc = accuracy_score(y_test, y_pred)\n",
    "print('Accuracy: ',acc)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Showing the results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x183b67cc128>"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import seaborn as sn\n",
    "\n",
    "df_cm = pd.DataFrame(cm, range(cm.shape[0]),\n",
    "                  range(cm.shape[1]))\n",
    "#plt.figure(figsize = (10,7))\n",
    "sn.set(font_scale=1.4) #for label size\n",
    "sn.heatmap(df_cm, annot=True,fmt='.2f',annot_kws={\"size\": 10})# font size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYoAAAEWCAYAAAB42tAoAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nnMuXJwrnnHP58kThnHMuX/8P3k79dzDBt1EAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# ROC Curve\n",
    "# plot no skill\n",
    "plt.title('Receiver Operating Characteristic')\n",
    "plt.plot(fpr, tpr, 'b', label = 'AUC = %0.2f' % roc_auc)\n",
    "plt.legend(loc = 'lower right')\n",
    "plt.plot([0, 1], [0, 1],'r--')\n",
    "plt.xlim([0, 1])\n",
    "plt.ylim([0, 1])\n",
    "plt.ylabel('True Positive Rate')\n",
    "plt.xlabel('False Positive Rate')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Building and training a XGBoosted Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from xgboost import XGBClassifier\n",
    "from sklearn.metrics import accuracy_score"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Simple train and score "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 74.87%\n"
     ]
    }
   ],
   "source": [
    "model = XGBClassifier()\n",
    "model.fit(X_train, y_train)\n",
    "# make predictions for test data\n",
    "y_pred = model.predict(X_test)\n",
    "predictions = [round(value) for value in y_pred]\n",
    "# evaluate predictions\n",
    "accuracy = accuracy_score(y_test, predictions)\n",
    "print(\"Accuracy: %.2f%%\" % (accuracy * 100.0))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Use early stopping to prevent overfiting"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0]\tvalidation_0-mlogloss:1.81184\n",
      "Will train until validation_0-mlogloss hasn't improved in 10 rounds.\n",
      "[1]\tvalidation_0-mlogloss:1.70536\n",
      "[2]\tvalidation_0-mlogloss:1.60817\n",
      "[3]\tvalidation_0-mlogloss:1.52564\n",
      "[4]\tvalidation_0-mlogloss:1.45405\n",
      "[5]\tvalidation_0-mlogloss:1.39166\n",
      "[6]\tvalidation_0-mlogloss:1.33615\n",
      "[7]\tvalidation_0-mlogloss:1.28701\n",
      "[8]\tvalidation_0-mlogloss:1.24252\n",
      "[9]\tvalidation_0-mlogloss:1.20185\n",
      "[10]\tvalidation_0-mlogloss:1.16566\n",
      "[11]\tvalidation_0-mlogloss:1.13161\n",
      "[12]\tvalidation_0-mlogloss:1.10187\n",
      "[13]\tvalidation_0-mlogloss:1.07309\n",
      "[14]\tvalidation_0-mlogloss:1.04709\n",
      "[15]\tvalidation_0-mlogloss:1.02314\n",
      "[16]\tvalidation_0-mlogloss:1.00087\n",
      "[17]\tvalidation_0-mlogloss:0.979585\n",
      "[18]\tvalidation_0-mlogloss:0.961162\n",
      "[19]\tvalidation_0-mlogloss:0.944345\n",
      "[20]\tvalidation_0-mlogloss:0.927887\n",
      "[21]\tvalidation_0-mlogloss:0.912448\n",
      "[22]\tvalidation_0-mlogloss:0.897978\n",
      "[23]\tvalidation_0-mlogloss:0.884746\n",
      "[24]\tvalidation_0-mlogloss:0.872194\n",
      "[25]\tvalidation_0-mlogloss:0.86135\n",
      "[26]\tvalidation_0-mlogloss:0.850859\n",
      "[27]\tvalidation_0-mlogloss:0.840192\n",
      "[28]\tvalidation_0-mlogloss:0.830766\n",
      "[29]\tvalidation_0-mlogloss:0.821667\n",
      "[30]\tvalidation_0-mlogloss:0.81308\n",
      "[31]\tvalidation_0-mlogloss:0.805381\n",
      "[32]\tvalidation_0-mlogloss:0.797806\n",
      "[33]\tvalidation_0-mlogloss:0.79061\n",
      "[34]\tvalidation_0-mlogloss:0.784409\n",
      "[35]\tvalidation_0-mlogloss:0.777697\n",
      "[36]\tvalidation_0-mlogloss:0.771665\n",
      "[37]\tvalidation_0-mlogloss:0.766238\n",
      "[38]\tvalidation_0-mlogloss:0.760811\n",
      "[39]\tvalidation_0-mlogloss:0.755335\n",
      "[40]\tvalidation_0-mlogloss:0.750256\n",
      "[41]\tvalidation_0-mlogloss:0.745287\n",
      "[42]\tvalidation_0-mlogloss:0.740927\n",
      "[43]\tvalidation_0-mlogloss:0.73663\n",
      "[44]\tvalidation_0-mlogloss:0.732068\n",
      "[45]\tvalidation_0-mlogloss:0.728019\n",
      "[46]\tvalidation_0-mlogloss:0.724066\n",
      "[47]\tvalidation_0-mlogloss:0.720815\n",
      "[48]\tvalidation_0-mlogloss:0.717392\n",
      "[49]\tvalidation_0-mlogloss:0.713658\n",
      "[50]\tvalidation_0-mlogloss:0.710379\n",
      "[51]\tvalidation_0-mlogloss:0.707149\n",
      "[52]\tvalidation_0-mlogloss:0.703898\n",
      "[53]\tvalidation_0-mlogloss:0.70115\n",
      "[54]\tvalidation_0-mlogloss:0.697972\n",
      "[55]\tvalidation_0-mlogloss:0.695513\n",
      "[56]\tvalidation_0-mlogloss:0.693178\n",
      "[57]\tvalidation_0-mlogloss:0.690494\n",
      "[58]\tvalidation_0-mlogloss:0.688374\n",
      "[59]\tvalidation_0-mlogloss:0.685957\n",
      "[60]\tvalidation_0-mlogloss:0.683423\n",
      "[61]\tvalidation_0-mlogloss:0.680923\n",
      "[62]\tvalidation_0-mlogloss:0.678737\n",
      "[63]\tvalidation_0-mlogloss:0.676557\n",
      "[64]\tvalidation_0-mlogloss:0.674479\n",
      "[65]\tvalidation_0-mlogloss:0.672227\n",
      "[66]\tvalidation_0-mlogloss:0.669988\n",
      "[67]\tvalidation_0-mlogloss:0.668124\n",
      "[68]\tvalidation_0-mlogloss:0.666232\n",
      "[69]\tvalidation_0-mlogloss:0.664284\n",
      "[70]\tvalidation_0-mlogloss:0.662345\n",
      "[71]\tvalidation_0-mlogloss:0.660581\n",
      "[72]\tvalidation_0-mlogloss:0.659151\n",
      "[73]\tvalidation_0-mlogloss:0.657571\n",
      "[74]\tvalidation_0-mlogloss:0.655342\n",
      "[75]\tvalidation_0-mlogloss:0.653611\n",
      "[76]\tvalidation_0-mlogloss:0.652269\n",
      "[77]\tvalidation_0-mlogloss:0.651073\n",
      "[78]\tvalidation_0-mlogloss:0.64932\n",
      "[79]\tvalidation_0-mlogloss:0.648034\n",
      "[80]\tvalidation_0-mlogloss:0.646556\n",
      "[81]\tvalidation_0-mlogloss:0.644651\n",
      "[82]\tvalidation_0-mlogloss:0.643441\n",
      "[83]\tvalidation_0-mlogloss:0.641236\n",
      "[84]\tvalidation_0-mlogloss:0.640314\n",
      "[85]\tvalidation_0-mlogloss:0.638693\n",
      "[86]\tvalidation_0-mlogloss:0.636962\n",
      "[87]\tvalidation_0-mlogloss:0.635491\n",
      "[88]\tvalidation_0-mlogloss:0.633852\n",
      "[89]\tvalidation_0-mlogloss:0.632589\n",
      "[90]\tvalidation_0-mlogloss:0.63175\n",
      "[91]\tvalidation_0-mlogloss:0.630481\n",
      "[92]\tvalidation_0-mlogloss:0.629176\n",
      "[93]\tvalidation_0-mlogloss:0.627693\n",
      "[94]\tvalidation_0-mlogloss:0.626395\n",
      "[95]\tvalidation_0-mlogloss:0.624847\n",
      "[96]\tvalidation_0-mlogloss:0.623857\n",
      "[97]\tvalidation_0-mlogloss:0.622736\n",
      "[98]\tvalidation_0-mlogloss:0.621107\n",
      "[99]\tvalidation_0-mlogloss:0.619613\n",
      "Accuracy: 74.87%\n"
     ]
    }
   ],
   "source": [
    "# fit model on training data\n",
    "model = XGBClassifier()\n",
    "eval_set = [(X_test, y_test)]\n",
    "model.fit(X_train, y_train, early_stopping_rounds=10, eval_metric=\"mlogloss\", eval_set=eval_set, verbose=True)\n",
    "# make predictions for test data\n",
    "y_pred = model.predict(X_test)\n",
    "predictions = [round(value) for value in y_pred]\n",
    "# evaluate predictions\n",
    "accuracy = accuracy_score(y_test, predictions)\n",
    "print(\"Accuracy: %.2f%%\" % (accuracy * 100.0))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.11665306 0.00693619 0.00625086 0.02033894 0.01081294 0.01420751\n",
      " 0.02833169 0.01276159 0.00985048 0.01355531 0.02229558 0.00711938\n",
      " 0.0211423  0.0933722  0.0066583  0.01004208 0.05041707 0.02392996\n",
      " 0.01250993 0.01098667 0.         0.04807513 0.00604195 0.0339755\n",
      " 0.02577008 0.00442801 0.00190394 0.01292626 0.         0.\n",
      " 0.00368925 0.         0.02352204 0.01841108 0.00862217 0.\n",
      " 0.         0.         0.02629612 0.19710079 0.00738642 0.01835677\n",
      " 0.01105236 0.00128078 0.00456985 0.         0.         0.01546755\n",
      " 0.02246774 0.01048422]\n"
     ]
    }
   ],
   "source": [
    "print(model.feature_importances_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Figure size 720x1440 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from xgboost import plot_importance\n",
    "plt.figure(figsize=(10,20))\n",
    "plot_importance(model)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Hyperparameter tunning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 10 folds for each of 16 candidates, totalling 160 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 8 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  34 tasks      | elapsed:  3.8min\n",
      "[Parallel(n_jobs=-1)]: Done 160 out of 160 | elapsed: 31.3min finished\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.model_selection import StratifiedKFold\n",
    "\n",
    "model = XGBClassifier()\n",
    "\n",
    "n_estimators = [50, 100, 150, 200]\n",
    "max_depth = [2, 4, 6, 8]\n",
    "param_grid = dict(max_depth=max_depth, n_estimators=n_estimators)\n",
    "\n",
    "kfold = StratifiedKFold(n_splits=10, shuffle=True, random_state=7)\n",
    "grid_search = GridSearchCV(model, param_grid, scoring=\"neg_log_loss\", n_jobs=-1, cv=kfold, verbose=1)\n",
    "result = grid_search.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best: -0.368889 using {'max_depth': 8, 'n_estimators': 200}\n",
      "-0.792418 (0.019440) with: {'max_depth': 2, 'n_estimators': 50}\n",
      "-0.682119 (0.021245) with: {'max_depth': 2, 'n_estimators': 100}\n",
      "-0.633146 (0.021427) with: {'max_depth': 2, 'n_estimators': 150}\n",
      "-0.602118 (0.021040) with: {'max_depth': 2, 'n_estimators': 200}\n",
      "-0.642494 (0.020331) with: {'max_depth': 4, 'n_estimators': 50}\n",
      "-0.543494 (0.020418) with: {'max_depth': 4, 'n_estimators': 100}\n",
      "-0.495503 (0.019557) with: {'max_depth': 4, 'n_estimators': 150}\n",
      "-0.468006 (0.020418) with: {'max_depth': 4, 'n_estimators': 200}\n",
      "-0.539366 (0.019385) with: {'max_depth': 6, 'n_estimators': 50}\n",
      "-0.454329 (0.020631) with: {'max_depth': 6, 'n_estimators': 100}\n",
      "-0.417578 (0.024626) with: {'max_depth': 6, 'n_estimators': 150}\n",
      "-0.397278 (0.024947) with: {'max_depth': 6, 'n_estimators': 200}\n",
      "-0.473768 (0.020015) with: {'max_depth': 8, 'n_estimators': 50}\n",
      "-0.402162 (0.022094) with: {'max_depth': 8, 'n_estimators': 100}\n",
      "-0.378367 (0.025373) with: {'max_depth': 8, 'n_estimators': 150}\n",
      "-0.368889 (0.027516) with: {'max_depth': 8, 'n_estimators': 200}\n"
     ]
    }
   ],
   "source": [
    "print(\"Best: %f using %s\" % (result.best_score_, result.best_params_))\n",
    "means = result.cv_results_['mean_test_score']\n",
    "stds = result.cv_results_['std_test_score']\n",
    "params = result.cv_results_['params']\n",
    "\n",
    "for mean, stdev, param in zip(means, stds, params):\n",
    "    print(\"%f (%f) with: %r\" % (mean, stdev, param))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The scikit-learn version is 0.20.3.\n"
     ]
    }
   ],
   "source": [
    "import sklearn\n",
    "print('The scikit-learn version is {}.'.format(sklearn.__version__))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "ename": "ModuleNotFoundError",
     "evalue": "No module named 'lightgbm'",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-1-d1e428c0cb5b>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[1;32mfrom\u001b[0m \u001b[0mlightgbm\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mLGBMModel\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mLGBMClassifier\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;31mModuleNotFoundError\u001b[0m: No module named 'lightgbm'"
     ]
    }
   ],
   "source": [
    "from lightgbm import LGBMModel,LGBMClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}