soln.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The autoreload extension is already loaded. To reload it, use:\n",
      "  %reload_ext autoreload\n"
     ]
    }
   ],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 1 Imports\n",
    "\n",
    "from sklearn.tree import DecisionTreeRegressor\n",
    "from sklearn.ensemble import RandomForestRegressor\n",
    "from sklearn.ensemble import GradientBoostingRegressor\n",
    "from sklearn.ensemble import AdaBoostRegressor\n",
    "from sklearn.datasets import load_boston\n",
    "from sklearn.model_selection import train_test_split, cross_val_score\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.metrics import mean_squared_error, r2_score\n",
    "from sklearn.ensemble.partial_dependence import plot_partial_dependence\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "%matplotlib inline\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "\n",
    "import textwrap\n",
    "from individual import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 2 Import Data\n",
    "\n",
    "(X_train, X_test, y_train, y_test), column_names = load_and_split_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 3 Define Models\n",
    "\n",
    "rf = RandomForestRegressor(n_estimators=100,\n",
    "                            n_jobs=-1,\n",
    "                            random_state=1)\n",
    "\n",
    "gdbr = GradientBoostingRegressor(learning_rate=0.1,\n",
    "                                  loss='ls',\n",
    "                                  n_estimators=100,\n",
    "                                  random_state=1)\n",
    "\n",
    "abr = AdaBoostRegressor(DecisionTreeRegressor(),\n",
    "                         learning_rate=0.1,\n",
    "                         loss='linear',\n",
    "                         n_estimators=100,\n",
    "                         random_state=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RandomForestRegressor     Train CV | MSE: 9.837 | R2: 0.867\n",
      "GradientBoostingRegressor Train CV | MSE: 8.679 | R2: 0.883\n",
      "AdaBoostRegressor         Train CV | MSE: 9.361 | R2: 0.874\n"
     ]
    }
   ],
   "source": [
    "# 4 Cross Validate\n",
    "\n",
    "k = 10\n",
    "cross_val(rf, X_train, y_train, k) \n",
    "cross_val(gdbr, X_train, y_train, k) \n",
    "cross_val(abr, X_train, y_train, k);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "# The best model is currently GBR, however the hyperparameters are not tuned - so this situation will\n",
    "# change after a proper grid-search"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cross validation score of Gradient Boosting Regressor with lr = 0.1:\n",
      "GradientBoostingRegressor Train CV | MSE: 8.679 | R2: 0.883\n",
      "Cross validation score of Gradient Boosting Regressor with lr = 1.0:\n",
      "GradientBoostingRegressor Train CV | MSE: 21.850 | R2: 0.718\n"
     ]
    }
   ],
   "source": [
    "# 5 Gradient Boosting with LR = 1\n",
    "\n",
    "gdbr_lr1 = GradientBoostingRegressor(learning_rate=1.0, loss='ls',\n",
    "                                     n_estimators=100, random_state=1) \n",
    "print(\"Cross validation score of Gradient Boosting Regressor with lr = 0.1:\")\n",
    "cross_val(gdbr, X_train, y_train, k) \n",
    "print(\"Cross validation score of Gradient Boosting Regressor with lr = 1.0:\")\n",
    "cross_val(gdbr_lr1, X_train, y_train, k);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "# A learning rate of 1 has more than doubled MSE, also R2 is worse"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "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": [
    "# 6 Investigate effect of the number of iterations\n",
    "\n",
    "gbr_sm = GradientBoostingRegressor(learning_rate=0.1, loss='ls', n_estimators=100, random_state=1)\n",
    "gbr_lg = GradientBoostingRegressor(learning_rate=1.0, loss='ls', n_estimators=100, random_state=1)\n",
    "\n",
    "stage_score_plot(gbr_sm, X_train, y_train, X_test, y_test)\n",
    "stage_score_plot(gbr_lg, X_train, y_train, X_test, y_test)\n",
    "plt.legend();\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 7 Investigate effect of learning rate\n",
    "\n",
    "# See result in previous example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 8 Explain plot behavior\n",
    "\n",
    "# For both learning rates the train error drops to near zero with\n",
    "# progressive boosting stages (iterations).  However, the larger learning\n",
    "# rate test error doesn't decrease as more iterations are added.  The larger\n",
    "# learning rate makes large changes per iteration and so likely overfits\n",
    "# the data.  For the lower learning rate model to get low test error, many\n",
    "# boosting stages are required."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "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": [
    "# 9 Compare the Gradient Boosting Regressor with lr = 0.1 to a Random Forest Regressor\n",
    "\n",
    "stage_score_plot(gdbr, X_train, y_train, X_test, y_test)\n",
    "rf_score_plot(rf, X_train, y_train, X_test, y_test)\n",
    "plt.legend();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Gradient Boost outperforms Random Forest at about 60 iterations."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "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": [
    "# 10 Compare Adaboost Regressor with lr = 0.1 to the Random Forest Regressor\n",
    "\n",
    "stage_score_plot(abr, X_train, y_train, X_test, y_test)\n",
    "rf_score_plot(rf, X_train, y_train, X_test, y_test)\n",
    "plt.legend();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [],
   "source": [
    "# With the current hyperparameters, Adaboost performs erratically.\n",
    "# With very few iterations it outperforms Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 288 candidates, totalling 864 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 448 tasks      | elapsed:    4.3s\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "best parameters: {'bootstrap': True, 'max_depth': None, 'max_features': None, 'min_samples_leaf': 1, 'min_samples_split': 2, 'n_estimators': 40, 'random_state': 1}\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Done 864 out of 864 | elapsed:    8.0s finished\n"
     ]
    }
   ],
   "source": [
    "# 11 Grid Search Random Forest\n",
    "\n",
    "random_forest_grid = {'max_depth': [3, None],\n",
    "                      'max_features': ['sqrt', 'log2', None],\n",
    "                      'min_samples_split': [2, 4],\n",
    "                      'min_samples_leaf': [1, 2, 4],\n",
    "                      'bootstrap': [True, False],\n",
    "                      'n_estimators': [10, 20, 40, 80],\n",
    "                      'random_state': [1]}\n",
    "\n",
    "rf_gridsearch = GridSearchCV(RandomForestRegressor(),\n",
    "                             random_forest_grid,\n",
    "                             n_jobs=-1,\n",
    "                             verbose=True,\n",
    "                             scoring='neg_mean_squared_error')\n",
    "rf_gridsearch.fit(X_train, y_train)\n",
    "\n",
    "print( \"best parameters:\", rf_gridsearch.best_params_ )\n",
    "\n",
    "best_rf_model = rf_gridsearch.best_estimator_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "8.740998774509805"
      ]
     },
     "execution_count": 72,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# mse for grid searched random forest\n",
    "mean_squared_error(y_test, best_rf_model.predict(X_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "8.945236166666664"
      ]
     },
     "execution_count": 73,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# mse for default random forest\n",
    "mean_squared_error(y_test, rf.predict(X_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "# grid search is better"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 144 candidates, totalling 432 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=1)]: Using backend SequentialBackend with 1 concurrent workers.\n",
      "[Parallel(n_jobs=1)]: Done 432 out of 432 | elapsed:   54.7s finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "best parameters: {'learning_rate': 0.05, 'max_depth': 4, 'max_features': 1.0, 'min_samples_leaf': 1, 'n_estimators': 500, 'random_state': 1}\n"
     ]
    }
   ],
   "source": [
    "# 12 Grid Search Gradient Boosting\n",
    "\n",
    "\n",
    "gradient_boosting_grid = {'learning_rate': [0.1, 0.05, 0.02, 0.01],\n",
    "                          'max_depth': [2, 4, 6],\n",
    "                          'min_samples_leaf': [1, 2, 5, 10],\n",
    "                          'max_features': [1.0, 0.3, 0.1],\n",
    "                          'n_estimators': [500],\n",
    "                          'random_state': [1]}\n",
    "\n",
    "\n",
    "gbr_gridsearch = GridSearchCV(GradientBoostingRegressor(),\n",
    "                             gradient_boosting_grid,\n",
    "                             verbose=True,\n",
    "                             scoring='neg_mean_squared_error')\n",
    "gbr_gridsearch.fit(X_train, y_train)\n",
    "\n",
    "print( \"best parameters:\", gbr_gridsearch.best_params_ )\n",
    "\n",
    "best_gbr_model = gbr_gridsearch.best_estimator_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "8.166498387403717"
      ]
     },
     "execution_count": 76,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# mse for grid searched gbr\n",
    "mean_squared_error(y_test, best_gbr_model.predict(X_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "8.221228849448858"
      ]
     },
     "execution_count": 77,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# mse for default gbr\n",
    "mean_squared_error(y_test, gdbr.predict(X_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [],
   "source": [
    "# grid searched gradient boosted regressor FTW!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}