Hybrid of regression trees and linear regression_19.03.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.datasets import load_digits\n",
    "\n",
    "# basic liraries\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "# plotting\n",
    "import matplotlib.pyplot as plt\n",
    "import plotly.plotly as py\n",
    "%matplotlib inline\n",
    "\n",
    "# sklearn\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.linear_model import Lasso, LinearRegression \n",
    "from sklearn.tree import DecisionTreeRegressor\n",
    "from sklearn.ensemble import RandomForestRegressor, ExtraTreesRegressor\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.model_selection import train_test_split, GridSearchCV, cross_val_predict, cross_val_score, cross_validate\n",
    "from sklearn.model_selection import learning_curve, train_test_split, KFold\n",
    "from sklearn.metrics import make_scorer, mean_squared_error\n",
    "\n",
    "from hybrid_linear_tree import *\n",
    "# from hybrid_tree import *\n",
    "from sklearn.ensemble import BaggingRegressor\n",
    "\n",
    "import os, re\n",
    "import time\n",
    "import pickle\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 0. Dataset (PIK Digital day)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset_name = 'pik_train'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_pik = pd.read_csv(dataset_name + '.csv', encoding='cp1251')\n",
    "\n",
    "columns = [\n",
    "    'id', 'id_bulk', 'spalen', 'date1', 'start_square', 'value', 'price', \n",
    "    'mean_sq', 'mean_fl', 'plan_s', 'plan_m', 'plan_l', 'vid_0', 'vid_1', \n",
    "    'vid_2', 'month', 'month_cnt', 'class_objecta', 'kolichestvo_pomecheniy', \n",
    "    'ogorojena_territoriya', 'plochad_zemelnogo_uchastka', 'vhodnie_gruppi',\n",
    "    'detskiy_sad', 'school', 'hospital', 'fok', 'sportivnaya_plochadka',\n",
    "    'avtomoika', 'kladovie', 'kolyasochnie', 'kondei', 'ventilaciya', 'lift',\n",
    "    'sistema_musora_otvedeniya', 'videonabludenie', 'podzemnaiya_parkovka',\n",
    "    'dvor_bez_machin', 'machinomesta', 'plochad_promzoni_radius_500_m',\n",
    "    'plochad_zelenoy_zoni_radius_500_m', 'do_kremliya_km', 'do_ttk_km',\n",
    "    'do_sadovogo_km', 'do_bolchoy_dorogi_na_machine',\n",
    "    'do_udobnoy_avtorazviyazki_km', 'do_metro_pechkom_km', 'do_promki_km',\n",
    "    'do_parka_km', 'do_parka_pechkom_km','stanciy_metro_ot_kolca',\n",
    "    'plochad_dvora', 'kurs', 'stavka_po_ipoteke', 'vkladi_ot_1_goda',\n",
    "    'vkladi_ot_1do_3_let', 'vkladi_sviche_3_let'\n",
    "]\n",
    "df_pik.columns = columns\n",
    "\n",
    "# Ненужные или Нет в тестовых данных\n",
    "df_pik.drop(['id', 'sistema_musora_otvedeniya', 'start_square', 'plan_s', 'plan_m', 'plan_l', 'vid_0', 'vid_1', 'vid_2'], \n",
    "        axis=1, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# convert string columns to binary\n",
    "\n",
    "df_pik['econom'] = 0\n",
    "df_pik['comfort'] = 0\n",
    "df_pik['standart'] = 0\n",
    "\n",
    "for index, row in df_pik.iterrows():\n",
    "    class_obj = row['class_objecta']\n",
    "    if class_obj == 'комфорт':\n",
    "        df_pik.loc[index, 'comfort'] = 1\n",
    "    elif class_obj == 'эконом':\n",
    "        df_pik.loc[index, 'econom'] = 1\n",
    "    else:\n",
    "        df_pik.loc[index, 'standart'] = 1\n",
    "\n",
    "\n",
    "# 2. Заменим \"да\" и \"нет\" на 1 и 0 соответсвенно в столбцах ogorojena_territoriya, avtomoika, kladovie, \n",
    "# kolyasochnie, podzemnaiya_parkovka, dvor_bez_machin, sportivnaya_plochadka и vhodnie_gruppi.\n",
    "\n",
    "for col in ['ogorojena_territoriya', 'avtomoika', 'kladovie'\\\n",
    "            ,'kolyasochnie', 'podzemnaiya_parkovka', 'dvor_bez_machin', 'sportivnaya_plochadka', 'vhodnie_gruppi']:\n",
    "    df_pik[col] = df_pik[col].apply(lambda elem: 1 if elem == 'да' else 0)\n",
    "\n",
    "df_pik.drop('class_objecta', axis = 1, inplace = True)\n",
    "\n",
    "df_pik.set_index(['date1', 'id_bulk'], inplace = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = df_pik.drop('value', axis = 1)\n",
    "y = df_pik['value']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(8726, 46)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Pipeline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.1 Lasso"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train MSE: 109941.22 +/- 3745.85\n",
      "CV MSE: 111229.38 +/- 15783.3\n",
      "{'model__alpha': 1}\n",
      "time: 0.05 min\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "cv = KFold(n_splits = 5, shuffle = True, random_state = 42)  # лучше cv = 10\n",
    "\n",
    "scl = StandardScaler(with_mean = True, with_std = True)\n",
    "model = Lasso(max_iter = 5000, tol = 1e-3, random_state= 42)\n",
    "\n",
    "pipe = Pipeline(\n",
    "    steps = [\n",
    "         ('scl', scl),\n",
    "         ('model', model)\n",
    "    ]\n",
    ")\n",
    "\n",
    "param_grid = { \n",
    "    'model__alpha': [1, 5, 10]\n",
    "}\n",
    "\n",
    "gs_lasso = GridSearchCV(\n",
    "    pipe, \n",
    "    param_grid,\n",
    "    cv = cv,\n",
    "    scoring = 'neg_mean_squared_error'\n",
    ")\n",
    "\n",
    "gs_lasso.fit(X, y)\n",
    "\n",
    "train_MSE = (-gs_lasso.cv_results_['mean_train_score'][gs_lasso.cv_results_['rank_test_score'] == 1]).mean()\n",
    "train_std = (gs_lasso.cv_results_['std_train_score'][gs_lasso.cv_results_['rank_test_score'] == 1]).mean()\n",
    "\n",
    "cv_MSE = (-gs_lasso.cv_results_['mean_test_score'][gs_lasso.cv_results_['rank_test_score'] == 1]).mean()\n",
    "cv_std = (gs_lasso.cv_results_['std_test_score'][gs_lasso.cv_results_['rank_test_score'] == 1]).mean()\n",
    "\n",
    "print('train MSE: {} +/- {}'.format(np.round(train_MSE, 2), np.round(train_std, 2)))\n",
    "print('CV MSE: {} +/- {}'.format(np.round(cv_MSE, 2), np.round(cv_std, 2)))\n",
    "print(gs_lasso.best_params_)\n",
    "print('time: {} min'.format(np.round((time.time() - start_time)/60, 2))) \n",
    "\n",
    "filename = dataset_name + '_Lasso.sav'\n",
    "pickle.dump(gs_lasso, open(filename, 'wb'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.2 Decision Tree"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train MSE: 19271.28 +/- 879.54\n",
      "CV MSE: 93506.16 +/- 14661.06\n",
      "{'max_depth': 30, 'min_impurity_decrease': 3, 'min_samples_leaf': 3}\n",
      "time :0.03 min\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "model = DecisionTreeRegressor(criterion='mse', random_state = 42)\n",
    "\n",
    "cv = KFold(n_splits = 5, shuffle = True, random_state = 42)  # лучше cv = 10\n",
    "param_grid = {\n",
    "    'max_depth': [30, 40, 50],\n",
    "    'min_samples_leaf': [3],  # Минимальное число наблюдений в листе\n",
    "    'min_impurity_decrease' : [3],  # Минимальный прирост MSE при разделении в вершине на 2 ветки\n",
    "}\n",
    "gs_tree = GridSearchCV(model, param_grid, cv = cv, scoring = 'neg_mean_squared_error')\n",
    "gs_tree.fit(X, y)\n",
    "\n",
    "train_MSE = (-gs_tree.cv_results_['mean_train_score'][gs_tree.cv_results_['rank_test_score'] == 1]).mean()\n",
    "train_std = (gs_tree.cv_results_['std_train_score'][gs_tree.cv_results_['rank_test_score'] == 1]).mean()\n",
    "\n",
    "cv_MSE = (-gs_tree.cv_results_['mean_test_score'][gs_tree.cv_results_['rank_test_score'] == 1]).mean()\n",
    "cv_std = (gs_tree.cv_results_['std_test_score'][gs_tree.cv_results_['rank_test_score'] == 1]).mean()\n",
    "\n",
    "print('train MSE: {} +/- {}'.format(np.round(train_MSE, 2), np.round(train_std, 2)))\n",
    "print('CV MSE: {} +/- {}'.format(np.round(cv_MSE, 2), np.round(cv_std, 2)))\n",
    "print(gs_tree.best_params_)\n",
    "print('time :{} min'.format(np.round((time.time() - start_time)/60, 2))) \n",
    "\n",
    "filename = dataset_name + '_DecisionTree.sav'\n",
    "pickle.dump(gs_tree, open(filename, 'wb'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.3 Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train RMSE: 23616.16 +/- 877.3\n",
      "CV RMSE: 59241.94 +/- 9051.58\n",
      "{'max_depth': 30, 'max_features': 0.7, 'min_impurity_decrease': 3, 'min_samples_leaf': 3, 'n_estimators': 100}\n",
      "time :1.16 min\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "model = RandomForestRegressor(criterion='mse', random_state = 42)\n",
    "\n",
    "cv = KFold(n_splits = 5, shuffle = True, random_state = 42)  # лучше cv = 10\n",
    "param_grid = {\n",
    "    'n_estimators': [100],\n",
    "    'max_depth': [20, 30],\n",
    "    'min_samples_leaf': [3],  # Минимальное число наблюдений в листе\n",
    "    'min_impurity_decrease' : [3],  # Минимальный прирост MSE при разделении в вершине на 2 ветки\n",
    "    'max_features': [0.7]  # 1, 0.7 - доля от всех фичей\n",
    "}\n",
    "gs_rf = GridSearchCV(model, param_grid, cv = cv, scoring = 'neg_mean_squared_error')\n",
    "gs_rf.fit(X, y)\n",
    "\n",
    "train_MSE = (-gs_rf.cv_results_['mean_train_score'][gs_rf.cv_results_['rank_test_score'] == 1]).mean()\n",
    "train_std = (gs_rf.cv_results_['std_train_score'][gs_rf.cv_results_['rank_test_score'] == 1]).mean()\n",
    "\n",
    "cv_MSE = (-gs_rf.cv_results_['mean_test_score'][gs_rf.cv_results_['rank_test_score'] == 1]).mean()\n",
    "cv_std = (gs_rf.cv_results_['std_test_score'][gs_rf.cv_results_['rank_test_score'] == 1]).mean()\n",
    "\n",
    "print('train RMSE: {} +/- {}'.format(np.round(train_MSE, 2), np.round(train_std, 2)))\n",
    "print('CV RMSE: {} +/- {}'.format(np.round(cv_MSE, 2), np.round(cv_std, 2)))\n",
    "print(gs_rf.best_params_)\n",
    "print('time :{} min'.format(np.round((time.time() - start_time)/60, 2))) \n",
    "\n",
    "filename = dataset_name + '_RF.sav'\n",
    "pickle.dump(gs_rf, open(filename, 'wb'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.4  ExtraTreeRegressor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train MSE: 19856.22 +/- 986.25\n",
      "CV MSE: 56203.49 +/- 6918.45\n",
      "{'max_depth': 30, 'max_features': 0.7, 'min_impurity_decrease': 3, 'min_samples_leaf': 3, 'n_estimators': 100}\n",
      "time :0.51 min\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "model = ExtraTreesRegressor(criterion='mse',\n",
    "                              random_state = 42)\n",
    "\n",
    "cv = KFold(n_splits = 5, shuffle = True, random_state = 42)  # лучше cv = 10\n",
    "param_grid = {\n",
    "             'n_estimators': [100],\n",
    "             'max_depth': [20, 30],\n",
    "             'min_samples_leaf': [3],  # Минимальное число наблюдений в листе\n",
    "             'min_impurity_decrease' : [3],  # Минимальный прирост MSE при разделении в вершине на 2 ветки\n",
    "             'max_features': [0.7]  # 1, 0.7 - доля от всех фичей\n",
    "             }\n",
    "gs_etr = GridSearchCV(model, param_grid,\n",
    "                 cv = cv,\n",
    "                 scoring = 'neg_mean_squared_error')\n",
    "gs_etr.fit(X, y)\n",
    "\n",
    "train_MSE = (-gs_etr.cv_results_['mean_train_score'][gs_etr.cv_results_['rank_test_score'] == 1]).mean()\n",
    "train_std = (gs_etr.cv_results_['std_train_score'][gs_etr.cv_results_['rank_test_score'] == 1]).mean()\n",
    "\n",
    "cv_MSE = (-gs_etr.cv_results_['mean_test_score'][gs_etr.cv_results_['rank_test_score'] == 1]).mean()\n",
    "cv_std = (gs_etr.cv_results_['std_test_score'][gs_etr.cv_results_['rank_test_score'] == 1]).mean()\n",
    "\n",
    "print('train MSE: {} +/- {}'.format(np.round(train_MSE, 2), np.round(train_std, 2)))\n",
    "print('CV MSE: {} +/- {}'.format(np.round(cv_MSE, 2), np.round(cv_std, 2)))\n",
    "print(gs_etr.best_params_)\n",
    "print('time :{} min'.format(np.round((time.time() - start_time)/60, 2))) \n",
    "\n",
    "filename = dataset_name + '_ExtraTreeRegressor.sav'\n",
    "pickle.dump(gs_etr, open(filename, 'wb'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.5 HRT"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "436"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "int(0.05*len(X))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train MSE: 63887.73 +/- 2060.34\n",
      "CV MSE: 85543.67 +/- 9519.01\n",
      "{'min_node_size': 261, 'min_split_improvement': 10, 'mode': 'Lasso'}\n",
      "time :7.77 min\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "model = HybridRegressionTree()\n",
    "\n",
    "cv = KFold(n_splits=5, shuffle=True, random_state=42)  # лучше cv = 10\n",
    "\n",
    "param_grid = {\n",
    "    'mode': ['Lasso'],\n",
    "    'min_node_size': [int(p*len(X)) for p in [0.03, 0.1, 0.3]],\n",
    "    'min_split_improvement': [10, 20]\n",
    "}\n",
    "\n",
    "gs_hrt = GridSearchCV(model, param_grid, cv=cv, scoring='neg_mean_squared_error')\n",
    "gs_hrt.fit(X, y.values)\n",
    "\n",
    "train_MSE = (-gs_hrt.cv_results_['mean_train_score'][gs_hrt.cv_results_['rank_test_score'] == 1]).mean()\n",
    "train_std = (gs_hrt.cv_results_['std_train_score'][gs_hrt.cv_results_['rank_test_score'] == 1]).mean()\n",
    "\n",
    "cv_MSE = (-gs_hrt.cv_results_['mean_test_score'][gs_hrt.cv_results_['rank_test_score'] == 1]).mean()\n",
    "cv_std = (gs_hrt.cv_results_['std_test_score'][gs_hrt.cv_results_['rank_test_score'] == 1]).mean()\n",
    "\n",
    "print('train MSE: {} +/- {}'.format(np.round(train_MSE, 2), np.round(train_std, 2)))\n",
    "print('CV MSE: {} +/- {}'.format(np.round(cv_MSE, 2), np.round(cv_std, 2)))\n",
    "print(gs_hrt.best_params_)\n",
    "print('time :{} min'.format(np.round((time.time() - start_time)/60, 2))) \n",
    "\n",
    "filename = dataset_name + '_HybridRegressionTree.sav'\n",
    "pickle.dump(gs_hrt, open(filename, 'wb'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "FUCK\n",
      "FUCK\n",
      "FUCK\n",
      "FUCK\n",
      "FUCK\n"
     ]
    }
   ],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "min_samples_node_list = [int(0.01*X.shape[0]), int(0.05*X.shape[0]),\\\n",
    "                         int(0.1*X.shape[0]), int(0.2*X.shape[0]), int(0.3*X.shape[0])]\n",
    "min_split_threshold_list = [10, 10, 10, 15, 20, 20]\n",
    "\n",
    "scores_hrt_df = pd.DataFrame(columns=['min_samples_leaf', 'min_impurity_decrease', \n",
    "                     'fold1', 'fold2', 'fold3', 'fold4', 'fold5'])\n",
    "\n",
    "\n",
    "for min_samples_node, min_split_threshold  in zip(min_samples_node_list, min_split_threshold_list):\n",
    "    \n",
    "    hrt = HybridRegressionTree(min_node_size = min_samples_node, \n",
    "                               mode='Lasso', \n",
    "                               min_split_improvement=min_split_threshold)\n",
    "    \n",
    "    kf = KFold(n_splits=5, random_state=42)\n",
    "    scores = []\n",
    "    for train_index, test_index in kf.split(X):\n",
    "        X_train, X_test = X.iloc[train_index], X.iloc[test_index]\n",
    "        y_train, y_test = y[train_index], y[test_index]\n",
    "\n",
    "        hrt.fit(X_train, y_train.values)\n",
    "        y_pred = hrt.predict(X_test)\n",
    "        rmse = np.sqrt(mean_squared_error(y_test, y_pred))\n",
    "        scores.append(rmse)\n",
    "        \n",
    "    scores_hrt_df_tmp = pd.DataFrame([[min_samples_node, min_split_threshold] + scores], \n",
    "                                 columns=['min_samples_leaf', 'min_impurity_decrease', \\\n",
    "                                             'fold1', 'fold2', 'fold3', 'fold4', 'fold5'])\n",
    "    scores_hrt_df = scores_hrt_df.append(scores_hrt_df_tmp)\n",
    "\n",
    "\n",
    "# cv_MSE = (-gs_hrt.cv_results_['mean_test_score'][gs_hrt.cv_results_['rank_test_score'] == 1]).mean()\n",
    "# cv_std = (gs_hrt.cv_results_['std_test_score'][gs_hrt.cv_results_['rank_test_score'] == 1]).mean()\n",
    "\n",
    "# print('train MSE: {} +/- {}'.format(np.round(train_MSE, 2), np.round(train_std, 2)))\n",
    "# print('CV MSE: {} +/- {}'.format(np.round(cv_MSE, 2), np.round(cv_std, 2)))\n",
    "# print(gs_hrt.best_params_)\n",
    "# print('time :{} min'.format(np.round((time.time() - start_time)/60, 2))) \n",
    "\n",
    "# filename = dataset_name + '_HRT.sav'\n",
    "# pickle.dump(gs_hrt, open(filename, 'wb'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "scores_hrt_df['RMSE'] = scores_hrt_df[['fold1', 'fold2', 'fold3', 'fold4', 'fold5']].mean(axis = 1).values\n",
    "scores_hrt_df['RMSE_std'] = scores_hrt_df[['fold1', 'fold2', 'fold3', 'fold4', 'fold5']].std(axis = 1).values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "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>min_samples_leaf</th>\n",
       "      <th>min_impurity_decrease</th>\n",
       "      <th>fold1</th>\n",
       "      <th>fold2</th>\n",
       "      <th>fold3</th>\n",
       "      <th>fold4</th>\n",
       "      <th>fold5</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>RMSE_std</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>87</td>\n",
       "      <td>10</td>\n",
       "      <td>1346.685369</td>\n",
       "      <td>1661.937938</td>\n",
       "      <td>1633.728466</td>\n",
       "      <td>899.004260</td>\n",
       "      <td>2064.193494</td>\n",
       "      <td>1521.109905</td>\n",
       "      <td>431.536387</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>436</td>\n",
       "      <td>10</td>\n",
       "      <td>686.511463</td>\n",
       "      <td>963.158077</td>\n",
       "      <td>560.069325</td>\n",
       "      <td>497.457725</td>\n",
       "      <td>665.892547</td>\n",
       "      <td>674.617828</td>\n",
       "      <td>178.871550</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>872</td>\n",
       "      <td>10</td>\n",
       "      <td>488.080476</td>\n",
       "      <td>674.336036</td>\n",
       "      <td>488.798118</td>\n",
       "      <td>402.646439</td>\n",
       "      <td>439.840677</td>\n",
       "      <td>498.740349</td>\n",
       "      <td>104.581950</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1745</td>\n",
       "      <td>15</td>\n",
       "      <td>484.134475</td>\n",
       "      <td>779.975666</td>\n",
       "      <td>565.066863</td>\n",
       "      <td>355.443104</td>\n",
       "      <td>313.818649</td>\n",
       "      <td>499.687751</td>\n",
       "      <td>186.031912</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2617</td>\n",
       "      <td>20</td>\n",
       "      <td>541.592654</td>\n",
       "      <td>668.912775</td>\n",
       "      <td>617.951539</td>\n",
       "      <td>347.812778</td>\n",
       "      <td>332.574286</td>\n",
       "      <td>501.768806</td>\n",
       "      <td>154.394557</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  min_samples_leaf min_impurity_decrease        fold1        fold2  \\\n",
       "0               87                    10  1346.685369  1661.937938   \n",
       "0              436                    10   686.511463   963.158077   \n",
       "0              872                    10   488.080476   674.336036   \n",
       "0             1745                    15   484.134475   779.975666   \n",
       "0             2617                    20   541.592654   668.912775   \n",
       "\n",
       "         fold3       fold4        fold5         RMSE    RMSE_std  \n",
       "0  1633.728466  899.004260  2064.193494  1521.109905  431.536387  \n",
       "0   560.069325  497.457725   665.892547   674.617828  178.871550  \n",
       "0   488.798118  402.646439   439.840677   498.740349  104.581950  \n",
       "0   565.066863  355.443104   313.818649   499.687751  186.031912  \n",
       "0   617.951539  347.812778   332.574286   501.768806  154.394557  "
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scores_hrt_df.head(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "hrt = HybridRegressionTree(mode='Lasso', min_node_size=2617, min_split_improvement=20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "def my_cross_validate(X, y, n_splits, hrt_params):\n",
    "    kf = KFold(n_splits=5, random_state=42)\n",
    "\n",
    "    hrt = HybridRegressionTree(**hrt_params)\n",
    "    \n",
    "    test_score = []\n",
    "    fit_time = []\n",
    "    \n",
    "    for train_index, test_index in kf.split(X):\n",
    "        print('ok')\n",
    "        X_train, X_test = X.iloc[train_index], X.iloc[test_index]\n",
    "        y_train, y_test = y[train_index], y[test_index]\n",
    "\n",
    "        t = time.time()\n",
    "        hrt.fit(X_train, y_train.values)\n",
    "        fit_time.append(time.time() - t)\n",
    "        \n",
    "        y_pred = hrt.predict(X_test)\n",
    "        test_score.append(np.sqrt(mean_squared_error(y_test, y_pred)))\n",
    "        \n",
    "    return dict(test_score=test_score, fit_time=fit_time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ok\n",
      "ok\n",
      "ok\n",
      "ok\n",
      "ok\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'fit_time': [1.6087064743041992,\n",
       "  1.5249207019805908,\n",
       "  1.570798397064209,\n",
       "  1.7403302192687988,\n",
       "  1.4650819301605225],\n",
       " 'test_score': [541.5926539554376,\n",
       "  668.9127750283317,\n",
       "  617.9515387342875,\n",
       "  347.81277808778134,\n",
       "  332.5742863245551]}"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "my_cross_validate(X, y, 5, dict(mode='Lasso', min_node_size=2617, min_split_improvement=20))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "min_samples_node_list = [int(0.01*X.shape[0]), int(0.05*X.shape[0]),\\\n",
    "                         int(0.1*X.shape[0]), int(0.2*X.shape[0]), int(0.3*X.shape[0])]\n",
    "min_split_threshold_list = [10, 10, 10, 15, 20, 20]\n",
    "\n",
    "scores_hrt_df = pd.DataFrame(columns=['min_samples_leaf', 'min_impurity_decrease', \n",
    "                     'fold1', 'fold2', 'fold3', 'fold4', 'fold5'])\n",
    "\n",
    "\n",
    "# for min_samples_node, min_split_threshold  in zip(min_samples_node_list, min_split_threshold_list):\n",
    "for min_samples_node, min_split_threshold  in zip([2617], [20]):\n",
    "    \n",
    "    hrt = HybridRegressionTree(min_node_size = min_samples_node, \n",
    "                               mode='Lasso', \n",
    "                               min_split_improvement=min_split_threshold)\n",
    "    \n",
    "    kf = KFold(n_splits=5, random_state=42)\n",
    "    scores = []\n",
    "    for train_index, test_index in kf.split(X):\n",
    "        X_train, X_test = X.iloc[train_index], X.iloc[test_index]\n",
    "        y_train, y_test = y[train_index], y[test_index]\n",
    "\n",
    "        hrt.fit(X_train, y_train.values)\n",
    "        y_pred = hrt.predict(X_test)\n",
    "        rmse = np.sqrt(mean_squared_error(y_test, y_pred))\n",
    "        scores.append(rmse)\n",
    "        \n",
    "    scores_hrt_df_tmp = pd.DataFrame([[min_samples_node, min_split_threshold] + scores], \n",
    "                                 columns=['min_samples_leaf', 'min_impurity_decrease', \\\n",
    "                                             'fold1', 'fold2', 'fold3', 'fold4', 'fold5'])\n",
    "    scores_hrt_df = scores_hrt_df.append(scores_hrt_df_tmp)\n",
    "\n",
    "\n",
    "# cv_MSE = (-gs_hrt.cv_results_['mean_test_score'][gs_hrt.cv_results_['rank_test_score'] == 1]).mean()\n",
    "# cv_std = (gs_hrt.cv_results_['std_test_score'][gs_hrt.cv_results_['rank_test_score'] == 1]).mean()\n",
    "\n",
    "# print('train MSE: {} +/- {}'.format(np.round(train_MSE, 2), np.round(train_std, 2)))\n",
    "# print('CV MSE: {} +/- {}'.format(np.round(cv_MSE, 2), np.round(cv_std, 2)))\n",
    "# print(gs_hrt.best_params_)\n",
    "# print('time :{} min'.format(np.round((time.time() - start_time)/60, 2))) \n",
    "\n",
    "# filename = dataset_name + '_HRT.sav'\n",
    "# pickle.dump(gs_hrt, open(filename, 'wb'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "scores_hrt_df['RMSE'] = scores_hrt_df[['fold1', 'fold2', 'fold3', 'fold4', 'fold5']].mean(axis = 1).values\n",
    "scores_hrt_df['RMSE_std'] = scores_hrt_df[['fold1', 'fold2', 'fold3', 'fold4', 'fold5']].std(axis = 1).values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "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>min_samples_leaf</th>\n",
       "      <th>min_impurity_decrease</th>\n",
       "      <th>fold1</th>\n",
       "      <th>fold2</th>\n",
       "      <th>fold3</th>\n",
       "      <th>fold4</th>\n",
       "      <th>fold5</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>RMSE_std</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2617</td>\n",
       "      <td>20</td>\n",
       "      <td>541.592654</td>\n",
       "      <td>668.912775</td>\n",
       "      <td>617.951539</td>\n",
       "      <td>347.812778</td>\n",
       "      <td>332.574286</td>\n",
       "      <td>501.768806</td>\n",
       "      <td>154.394557</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  min_samples_leaf min_impurity_decrease       fold1       fold2       fold3  \\\n",
       "0             2617                    20  541.592654  668.912775  617.951539   \n",
       "\n",
       "        fold4       fold5        RMSE    RMSE_std  \n",
       "0  347.812778  332.574286  501.768806  154.394557  "
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scores_hrt_df.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "871; 10"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1.6 HRT Bagging"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "def bootstrap(X, y):\n",
    "    \n",
    "    size = X.shape[0]\n",
    "    idx_range = range(size)\n",
    "    new_idx = np.random.choice(idx_range, size, replace=True)\n",
    "    return X.iloc[new_idx], y.iloc[new_idx]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "start_time = time.time()\n",
    "\n",
    "min_samples_node_list = [int(0.05*X.shape[0]), int(0.1*X.shape[0]),\\\n",
    "                         int(0.1*X.shape[0]), int(0.1*X.shape[0]), int(0.2*X.shape[0])]\n",
    "min_split_threshold_list = [15, 15, 15, 15, 15, 15]\n",
    "n_estimators_list = [3, 4, 5, 6, 7, 3]\n",
    "\n",
    "scores_bag_hrt_df = pd.DataFrame(columns=['n_estimators', 'min_samples_leaf', 'min_impurity_decrease', \n",
    "                     'fold1', 'fold2', 'fold3', 'fold4', 'fold5'])\n",
    "\n",
    "\n",
    "for min_samples_node, min_split_threshold, n_estimators  in zip(min_samples_node_list, \n",
    "                                                                  min_split_threshold_list,\n",
    "                                                                  n_estimators_list):\n",
    "\n",
    "    \n",
    "    trees = []\n",
    "    for i in range(n_estimators):\n",
    "        trees.append(HybridRegressionTree(min_node_size = min_samples_node, \n",
    "                                          mode='Lasso', \n",
    "                                          min_split_improvement=min_split_threshold)\n",
    "                    )\n",
    "    \n",
    "    kf = KFold(n_splits=5, random_state=42)\n",
    "    scores = []\n",
    "    for train_index, test_index in kf.split(X):\n",
    "        \n",
    "        X_train, X_test = X.iloc[train_index], X.iloc[test_index]\n",
    "        y_train, y_test = y[train_index], y[test_index]\n",
    "        y_pred = np.zeros(len(test_index))\n",
    "        \n",
    "        for i in range(n_estimators):\n",
    "\n",
    "            X_train_new, y_train_new = bootstrap(X_train, y_train)\n",
    "            trees[i].fit(X_train_new.values, X_train_new, y_train_new.values)\n",
    "            y_pred = y_pred + trees[i].predict(X_test.values, X_test)\n",
    "        y_pred = y_pred / n_estimators   # averaging\n",
    "        rmse = np.sqrt(mean_squared_error(y_test, y_pred))\n",
    "        scores.append(rmse)\n",
    "        \n",
    "    scores_bag_hrt_df_tmp = pd.DataFrame([[n_estimators, min_samples_node, min_split_threshold] + scores], \n",
    "                                 columns=['n_estimators' ,'min_samples_leaf', 'min_impurity_decrease', \\\n",
    "                                             'fold1', 'fold2', 'fold3', 'fold4', 'fold5'])\n",
    "    scores_bag_hrt_df = scores_bag_hrt_df.append(scores_bag_hrt_df_tmp)\n",
    "\n",
    "\n",
    "# cv_MSE = (-gs_hrt.cv_results_['mean_test_score'][gs_hrt.cv_results_['rank_test_score'] == 1]).mean()\n",
    "# cv_std = (gs_hrt.cv_results_['std_test_score'][gs_hrt.cv_results_['rank_test_score'] == 1]).mean()\n",
    "\n",
    "# print('train MSE: {} +/- {}'.format(np.round(train_MSE, 2), np.round(train_std, 2)))\n",
    "# print('CV MSE: {} +/- {}'.format(np.round(cv_MSE, 2), np.round(cv_std, 2)))\n",
    "# print(gs_hrt.best_params_)\n",
    "# print('time :{} min'.format(np.round((time.time() - start_time)/60, 2))) \n",
    "\n",
    "# filename = dataset_name + '_HRT.sav'\n",
    "# pickle.dump(gs_hrt, open(filename, 'wb'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "scores_bag_hrt_df['RMSE'] = scores_bag_hrt_df[['fold1', 'fold2', 'fold3', 'fold4', 'fold5']].mean(axis = 1).values\n",
    "scores_bag_hrt_df['RMSE_std'] = scores_bag_hrt_df[['fold1', 'fold2', 'fold3', 'fold4', 'fold5']].std(axis = 1).values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "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>n_estimators</th>\n",
       "      <th>min_samples_leaf</th>\n",
       "      <th>min_impurity_decrease</th>\n",
       "      <th>fold1</th>\n",
       "      <th>fold2</th>\n",
       "      <th>fold3</th>\n",
       "      <th>fold4</th>\n",
       "      <th>fold5</th>\n",
       "      <th>RMSE</th>\n",
       "      <th>RMSE_std</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>436</td>\n",
       "      <td>15</td>\n",
       "      <td>516.536661</td>\n",
       "      <td>700.244914</td>\n",
       "      <td>562.947017</td>\n",
       "      <td>434.855131</td>\n",
       "      <td>583.612211</td>\n",
       "      <td>559.639187</td>\n",
       "      <td>97.191186</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>4</td>\n",
       "      <td>872</td>\n",
       "      <td>15</td>\n",
       "      <td>502.108814</td>\n",
       "      <td>517.501140</td>\n",
       "      <td>466.877963</td>\n",
       "      <td>366.126287</td>\n",
       "      <td>410.895217</td>\n",
       "      <td>452.701884</td>\n",
       "      <td>63.411515</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>5</td>\n",
       "      <td>872</td>\n",
       "      <td>15</td>\n",
       "      <td>521.772450</td>\n",
       "      <td>516.819554</td>\n",
       "      <td>473.069446</td>\n",
       "      <td>358.844905</td>\n",
       "      <td>406.402777</td>\n",
       "      <td>455.381827</td>\n",
       "      <td>71.087069</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6</td>\n",
       "      <td>872</td>\n",
       "      <td>15</td>\n",
       "      <td>485.671133</td>\n",
       "      <td>517.915946</td>\n",
       "      <td>480.255344</td>\n",
       "      <td>366.409602</td>\n",
       "      <td>385.804735</td>\n",
       "      <td>447.211352</td>\n",
       "      <td>66.839164</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>7</td>\n",
       "      <td>1745</td>\n",
       "      <td>15</td>\n",
       "      <td>520.318362</td>\n",
       "      <td>666.298226</td>\n",
       "      <td>556.877028</td>\n",
       "      <td>355.188404</td>\n",
       "      <td>351.960389</td>\n",
       "      <td>490.128482</td>\n",
       "      <td>135.740398</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  n_estimators min_samples_leaf min_impurity_decrease       fold1       fold2  \\\n",
       "0            3              436                    15  516.536661  700.244914   \n",
       "0            4              872                    15  502.108814  517.501140   \n",
       "0            5              872                    15  521.772450  516.819554   \n",
       "0            6              872                    15  485.671133  517.915946   \n",
       "0            7             1745                    15  520.318362  666.298226   \n",
       "\n",
       "        fold3       fold4       fold5        RMSE    RMSE_std  \n",
       "0  562.947017  434.855131  583.612211  559.639187   97.191186  \n",
       "0  466.877963  366.126287  410.895217  452.701884   63.411515  \n",
       "0  473.069446  358.844905  406.402777  455.381827   71.087069  \n",
       "0  480.255344  366.409602  385.804735  447.211352   66.839164  \n",
       "0  556.877028  355.188404  351.960389  490.128482  135.740398  "
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scores_bag_hrt_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "6; 872; 15"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Learning curves "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_learning_curve(estimators, labels, colors,\n",
    "                        title, \n",
    "                        X, y, \n",
    "                        ylim=None, \n",
    "                        cv=None, scoring = 'neg_mean_squared_error',\n",
    "                        n_jobs=None, \n",
    "                        train_sizes=np.linspace(.4, 1.0, 5)):\n",
    "    \"\"\"\n",
    "    Generate a simple plot of the test and training learning curve.\n",
    "\n",
    "    Parameters\n",
    "    ----------\n",
    "    estimators : list, objectS type that implements the \"fit\" and \"predict\" methods\n",
    "        An object of that type which is cloned for each validation.\n",
    "\n",
    "    title : string\n",
    "        Title for the chart.\n",
    "\n",
    "    X : array-like, shape (n_samples, n_features)\n",
    "        Training vector, where n_samples is the number of samples and\n",
    "        n_features is the number of features.\n",
    "\n",
    "    y : array-like, shape (n_samples) or (n_samples, n_features), optional\n",
    "        Target relative to X for classification or regression;\n",
    "        None for unsupervised learning.\n",
    "\n",
    "    ylim : tuple, shape (ymin, ymax), optional\n",
    "        Defines minimum and maximum yvalues plotted.\n",
    "\n",
    "    cv : int, cross-validation generator or an iterable, optional\n",
    "        Determines the cross-validation splitting strategy.\n",
    "        Possible inputs for cv are:\n",
    "          - None, to use the default 3-fold cross-validation,\n",
    "          - integer, to specify the number of folds.\n",
    "          - :term:`CV splitter`,\n",
    "          - An iterable yielding (train, test) splits as arrays of indices.\n",
    "\n",
    "        For integer/None inputs, if ``y`` is binary or multiclass,\n",
    "        :class:`StratifiedKFold` used. If the estimator is not a classifier\n",
    "        or if ``y`` is neither binary nor multiclass, :class:`KFold` is used.\n",
    "\n",
    "        Refer :ref:`User Guide <cross_validation>` for the various\n",
    "        cross-validators that can be used here.\n",
    "\n",
    "    n_jobs : int or None, optional (default=None)\n",
    "        Number of jobs to run in parallel.\n",
    "        ``None`` means 1 unless in a :obj:`joblib.parallel_backend` context.\n",
    "        ``-1`` means using all processors. See :term:`Glossary <n_jobs>`\n",
    "        for more details.\n",
    "\n",
    "    train_sizes : array-like, shape (n_ticks,), dtype float or int\n",
    "        Relative or absolute numbers of training examples that will be used to\n",
    "        generate the learning curve. If the dtype is float, it is regarded as a\n",
    "        fraction of the maximum size of the training set (that is determined\n",
    "        by the selected validation method), i.e. it has to be within (0, 1].\n",
    "        Otherwise it is interpreted as absolute sizes of the training sets.\n",
    "        Note that for classification the number of samples usually have to\n",
    "        be big enough to contain at least one sample from each class.\n",
    "        (default: np.linspace(0.1, 1.0, 5))\n",
    "    \"\"\"\n",
    "    plt.figure(figsize = (10, 6))\n",
    "    plt.title(title, fontsize = 16)\n",
    "    if ylim is not None:\n",
    "        plt.ylim(*ylim)\n",
    "    plt.xlabel(\"Training size\", fontsize = 14)\n",
    "    plt.ylabel(\"Negative RMSE\", fontsize = 14)\n",
    "    plt.grid()\n",
    "    \n",
    "    for estimator, label, color in zip(estimators, labels, colors):\n",
    "        train_sizes, train_scores, test_scores = learning_curve(\n",
    "            estimator, X, y, \n",
    "            cv=cv, scoring = scoring, \n",
    "            n_jobs=n_jobs, train_sizes=train_sizes)\n",
    "    \n",
    "        test_scores_mean = np.mean(test_scores, axis=1)\n",
    "        test_scores_std = np.std(test_scores, axis=1)\n",
    "        \n",
    "        plt.plot(np.round(train_sizes/X.shape[0], 1), test_scores_mean, 'o-', color= color,\n",
    "             label = label)\n",
    "    \n",
    "        plt.fill_between(np.round(train_sizes/X.shape[0], 1), test_scores_mean - test_scores_std,\n",
    "                         test_scores_mean + test_scores_std, alpha=0.1, color=color)\n",
    "    \n",
    "\n",
    "    plt.legend(loc=\"upper right\", fontsize = 12)\n",
    "    return plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "def rmse(y_true, y_pred): \n",
    "    return np.sqrt(mean_squared_error(y_true, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GridSearchCV\n",
      "GridSearchCV\n",
      "GridSearchCV\n",
      "GridSearchCV\n",
      "GridSearchCV\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<module 'matplotlib.pyplot' from 'C:\\\\Users\\\\mskuz\\\\Anaconda3\\\\lib\\\\site-packages\\\\matplotlib\\\\pyplot.py'>"
      ]
     },
     "execution_count": 58,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1f043193be0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "title = \"Learning Curves for best parameters, PIK dataset\"\n",
    "# Cross validation with 100 iterations to get smoother mean test and train\n",
    "# score curves, each time with 20% data randomly selected as a validation set.\n",
    "# cv = train_test_split(X, y, test_size=0.3, random_state=42, shuffle = True)\n",
    "cv = KFold(n_splits = 3, shuffle = True, random_state = 42)\n",
    "\n",
    "scoring = make_scorer(rmse)\n",
    "\n",
    "estimators = [gs_lasso, gs_tree, gs_rf, gs_etr, hrt_etr]\n",
    "labels = ['Lasso', 'Decision Tree', 'Random Forest', 'Extremely randomised RF', 'HybridRegressionTree']\n",
    "colors = ['red', 'green', 'blue', 'orange', 'purple']\n",
    "\n",
    "plot_learning_curve(estimators, labels, colors, title, X, y.values, ylim=(150, 1100), cv=cv, scoring = scoring, n_jobs=1)\n",
    "\n",
    "# Unfortunately 'learning_curve' function does not support non-sklearn like estimators. \n",
    "# So, we will add our methods to the plot manually"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Estimation time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_training_time(estimators, labels, colors,\n",
    "                        title, \n",
    "                        X, y, \n",
    "                        ylim=None, \n",
    "                        n_jobs=None, \n",
    "                        train_sizes=np.linspace(.4, 1.0, 5)):\n",
    "    \"\"\"\n",
    "    Generate a simple plot of the test and training learning curve.\n",
    "\n",
    "    Parameters\n",
    "    ----------\n",
    "    estimators : list, objectS type that implements the \"fit\" and \"predict\" methods\n",
    "        An object of that type which is cloned for each validation.\n",
    "\n",
    "    title : string\n",
    "        Title for the chart.\n",
    "\n",
    "    X : array-like, shape (n_samples, n_features)\n",
    "        Training vector, where n_samples is the number of samples and\n",
    "        n_features is the number of features.\n",
    "\n",
    "    y : array-like, shape (n_samples) or (n_samples, n_features), optional\n",
    "        Target relative to X for classification or regression;\n",
    "        None for unsupervised learning.\n",
    "\n",
    "    ylim : tuple, shape (ymin, ymax), optional\n",
    "        Defines minimum and maximum yvalues plotted.\n",
    "\n",
    "    n_jobs : int or None, optional (default=None)\n",
    "        Number of jobs to run in parallel.\n",
    "        ``None`` means 1 unless in a :obj:`joblib.parallel_backend` context.\n",
    "        ``-1`` means using all processors. See :term:`Glossary <n_jobs>`\n",
    "        for more details.\n",
    "\n",
    "    train_sizes : array-like, shape (n_ticks,), dtype float or int\n",
    "        Relative or absolute numbers of training examples that will be used to\n",
    "        generate the learning curve. If the dtype is float, it is regarded as a\n",
    "        fraction of the maximum size of the training set (that is determined\n",
    "        by the selected validation method), i.e. it has to be within (0, 1].\n",
    "        Otherwise it is interpreted as absolute sizes of the training sets.\n",
    "        Note that for classification the number of samples usually have to\n",
    "        be big enough to contain at least one sample from each class.\n",
    "        (default: np.linspace(0.1, 1.0, 5))\n",
    "    \"\"\"\n",
    "    \n",
    "    plt.figure(figsize = (10, 6))\n",
    "    plt.title(title, fontsize = 16)\n",
    "    if ylim is not None:\n",
    "        plt.ylim(*ylim)\n",
    "    plt.yscale(\"log\")\n",
    "    plt.xlabel(\"Train size\", fontsize = 14)\n",
    "    plt.ylabel(\"Time (seconds), log scale\", fontsize = 14)\n",
    "    plt.title(title, fontsize = 16)\n",
    "    plt.grid()\n",
    "\n",
    "    for label, estimator, color in zip(labels, estimators, colors):\n",
    "        train_time = []\n",
    "        test_time = []\n",
    "        for size in train_sizes:\n",
    "            \n",
    "            size = int(size * len(X))\n",
    "            t0 = time.time()\n",
    "            estimator.fit(X[:size], y[:size])\n",
    "            train_time.append(time.time() - t0)\n",
    "\n",
    "        plt.plot(train_sizes*X.shape[0], train_time, 'o-', color=color, label=label)\n",
    "\n",
    "\n",
    "    plt.legend(loc='lower right', fontsize = 12)\n",
    "    \n",
    "    return plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<module 'matplotlib.pyplot' from 'C:\\\\Users\\\\mskuz\\\\Anaconda3\\\\lib\\\\site-packages\\\\matplotlib\\\\pyplot.py'>"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1f04329dda0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "title = \"Training Time for best parameters, PIK dataset\"\n",
    "estimators = [gs_lasso, gs_tree, gs_rf, gs_etr, hrt_etr]\n",
    "labels = ['Lasso', 'Decision Tree', 'Random Forest', 'Extremely randomised RF', 'HybridRegressionTree']\n",
    "colors = ['red', 'green', 'blue', 'orange', 'purple']\n",
    "\n",
    "plot_training_time(estimators, labels, colors, title, X, y.values, n_jobs=1)"
   ]
  }
 ],
 "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": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}