Week 2/code/.ipynb_checkpoints/Linear Regression- More Advanced Techniques-checkpoint.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Learning Scikit-learn: Machine Learning in Python"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Estimating Boston house pricing using Linear Regression- More Advanced Techniques"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We want to solve a *regression* problem (we want to approximate a real function)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Import the Boston House Pricing Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(506, 13)\n",
      "['CRIM' 'ZN' 'INDUS' 'CHAS' 'NOX' 'RM' 'AGE' 'DIS' 'RAD' 'TAX' 'PTRATIO'\n",
      " 'B' 'LSTAT']\n",
      "50.0 5.0 22.5328063241\n",
      "Boston House Prices dataset\n",
      "===========================\n",
      "\n",
      "Notes\n",
      "------\n",
      "Data Set Characteristics:  \n",
      "\n",
      "    :Number of Instances: 506 \n",
      "\n",
      "    :Number of Attributes: 13 numeric/categorical predictive\n",
      "    \n",
      "    :Median Value (attribute 14) is usually the target\n",
      "\n",
      "    :Attribute Information (in order):\n",
      "        - CRIM     per capita crime rate by town\n",
      "        - ZN       proportion of residential land zoned for lots over 25,000 sq.ft.\n",
      "        - INDUS    proportion of non-retail business acres per town\n",
      "        - CHAS     Charles River dummy variable (= 1 if tract bounds river; 0 otherwise)\n",
      "        - NOX      nitric oxides concentration (parts per 10 million)\n",
      "        - RM       average number of rooms per dwelling\n",
      "        - AGE      proportion of owner-occupied units built prior to 1940\n",
      "        - DIS      weighted distances to five Boston employment centres\n",
      "        - RAD      index of accessibility to radial highways\n",
      "        - TAX      full-value property-tax rate per $10,000\n",
      "        - PTRATIO  pupil-teacher ratio by town\n",
      "        - B        1000(Bk - 0.63)^2 where Bk is the proportion of blacks by town\n",
      "        - LSTAT    % lower status of the population\n",
      "        - MEDV     Median value of owner-occupied homes in $1000's\n",
      "\n",
      "    :Missing Attribute Values: None\n",
      "\n",
      "    :Creator: Harrison, D. and Rubinfeld, D.L.\n",
      "\n",
      "This is a copy of UCI ML housing dataset.\n",
      "http://archive.ics.uci.edu/ml/datasets/Housing\n",
      "\n",
      "\n",
      "This dataset was taken from the StatLib library which is maintained at Carnegie Mellon University.\n",
      "\n",
      "The Boston house-price data of Harrison, D. and Rubinfeld, D.L. 'Hedonic\n",
      "prices and the demand for clean air', J. Environ. Economics & Management,\n",
      "vol.5, 81-102, 1978.   Used in Belsley, Kuh & Welsch, 'Regression diagnostics\n",
      "...', Wiley, 1980.   N.B. Various transformations are used in the table on\n",
      "pages 244-261 of the latter.\n",
      "\n",
      "The Boston house-price data has been used in many machine learning papers that address regression\n",
      "problems.   \n",
      "     \n",
      "**References**\n",
      "\n",
      "   - Belsley, Kuh & Welsch, 'Regression diagnostics: Identifying Influential Data and Sources of Collinearity', Wiley, 1980. 244-261.\n",
      "   - Quinlan,R. (1993). Combining Instance-Based and Model-Based Learning. In Proceedings on the Tenth International Conference of Machine Learning, 236-243, University of Massachusetts, Amherst. Morgan Kaufmann.\n",
      "   - many more! (see http://archive.ics.uci.edu/ml/datasets/Housing)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.datasets import load_boston\n",
    "boston = load_boston()\n",
    "print(boston.data.shape)\n",
    "print(boston.feature_names)\n",
    "print(np.max(boston.target), np.min(boston.target), np.mean(boston.target))\n",
    "print(boston.DESCR)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[  6.32000000e-03   1.80000000e+01   2.31000000e+00   0.00000000e+00\n",
      "   5.38000000e-01   6.57500000e+00   6.52000000e+01   4.09000000e+00\n",
      "   1.00000000e+00   2.96000000e+02   1.53000000e+01   3.96900000e+02\n",
      "   4.98000000e+00]\n",
      "711.0 0.0 70.0724468258\n"
     ]
    }
   ],
   "source": [
    "print(boston.data[0])\n",
    "print(np.max(boston.data), np.min(boston.data), np.mean(boston.data))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Separate train and test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "X_train, X_test, y_train, y_test = train_test_split(boston.data, boston.target, test_size=0.25, random_state=33)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "711.0 0.0 69.6306540004 50.0 5.0 22.9237467018\n"
     ]
    }
   ],
   "source": [
    "print(np.max(X_train), np.min(X_train), np.mean(X_train), np.max(y_train), np.min(y_train), np.mean(y_train))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Find the most important features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<zip object at 0x10c960c48>\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAtEAAAD5CAYAAADhqWl1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nuYiWJEmSenIRLUmSJPXkIlqSJEnq6U8f26MIWDFn/QAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10c7d2d30>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.feature_selection import *\n",
    "fs=SelectKBest(score_func=f_regression,k=5)\n",
    "X_new=fs.fit_transform(X_train,y_train)\n",
    "print(zip(fs.get_support(),boston.feature_names))\n",
    "\n",
    "x_min, x_max = X_new[:,0].min() - .5, X_new[:, 0].max() + .5\n",
    "y_min, y_max = y_train.min() - .5, y_train.max() + .5\n",
    "#fig=plt.figure()\n",
    "#fig.subplots_adjust(left=0, right=1, bottom=0, top=1, hspace=0.05, wspace=0.05)\n",
    "\n",
    "# Two subplots, unpack the axes array immediately\n",
    "fig, axes = plt.subplots(1,5)\n",
    "fig.set_size_inches(12,12)\n",
    "\n",
    "for i in range(5):\n",
    "    axes[i].set_aspect('equal')\n",
    "    axes[i].set_title('Feature ' + str(i))\n",
    "    axes[i].set_xlabel('Feature')\n",
    "    axes[i].set_ylabel('Median house value')\n",
    "    axes[i].set_xlim(x_min, x_max)\n",
    "    axes[i].set_ylim(y_min, y_max)\n",
    "    plt.sca(axes[i])\n",
    "    plt.scatter(X_new[:,i],y_train)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Normalize data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10.2028980046 -4.66702040845 2.47038706385e-15 50.0 5.0 22.9237467018\n"
     ]
    }
   ],
   "source": [
    "from sklearn.preprocessing import StandardScaler\n",
    "X_train = StandardScaler().fit_transform(X_train)\n",
    "#scalerX = StandardScaler().fit(X_train)\n",
    "#scalery = StandardScaler().fit(y_train)\n",
    "\n",
    "#X_train = scalerX.transform(X_train)\n",
    "#y_train = StandardScaler().fit_transform((y_train)\n",
    "X_test = StandardScaler().fit_transform(X_test)\n",
    "#y_test = StandardScaler().fit_transform(y_test)\n",
    "\n",
    "print(np.max(X_train), np.min(X_train), np.mean(X_train), np.max(y_train), np.min(y_train), np.mean(y_train))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Start with a lineal model, SGDRegressor, that tries to find the hyperplane that minimizes a certain loss function (typically, the sum of squared distances from each instance to the hyperplane). It uses Stochastic Gradient Descent to find the minimum. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import *\n",
    "def train_and_evaluate(clf, X_train, y_train):\n",
    "    \n",
    "    clf.fit(X_train, y_train)\n",
    "    \n",
    "    print(\"Coefficient of determination on training set:\",clf.score(X_train, y_train))\n",
    "    \n",
    "    # create a k-fold croos validation iterator of k=5 folds\n",
    "    scores = cross_val_score(clf, X_train, y_train, cv=5)\n",
    "    print(\"Average coefficient of determination using 5-fold crossvalidation:\",np.mean(scores))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SGDRegressor(alpha=0.0001, average=False, epsilon=0.1, eta0=0.01,\n",
      "       fit_intercept=True, l1_ratio=0.15, learning_rate='invscaling',\n",
      "       loss='squared_loss', n_iter=5, penalty=None, power_t=0.25,\n",
      "       random_state=42, shuffle=False, verbose=0, warm_start=False)\n",
      "Coefficient of determination on training set: 0.740500682489\n",
      "Average coefficient of determination using 5-fold crossvalidation: 0.692832085398\n",
      "[-0.68040164  0.64814359 -0.5446693   1.01795136 -0.50131255  3.41563825\n",
      " -0.42909141 -1.92804105  0.74874927 -0.45816418 -1.85102023  0.6579521\n",
      " -3.44585513]\n"
     ]
    }
   ],
   "source": [
    "from sklearn import linear_model\n",
    "clf_sgd = linear_model.SGDRegressor(loss='squared_loss', penalty=None,  random_state=42, shuffle=False)\n",
    "print(clf_sgd)\n",
    "train_and_evaluate(clf_sgd,X_train,y_train)\n",
    "print(clf_sgd.coef_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf_sgd1 = linear_model.SGDRegressor(loss='squared_loss', penalty='l2',  random_state=42)\n",
    "train_and_evaluate(clf_sgd1,X_train,y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf_sgd2 = linear_model.SGDRegressor(loss='squared_loss', penalty='l1',  random_state=42)\n",
    "train_and_evaluate(clf_sgd2,X_train,y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf_sgd3 = linear_model.SGDRegressor(loss='squared_loss', penalty='elasticnet',  random_state=42)\n",
    "train_and_evaluate(clf_sgd3,X_train,y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ridge regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf_ridge = linear_model.Ridge()\n",
    "train_and_evaluate(clf_ridge,X_train,y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "SVR"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import svm\n",
    "clf_svr= svm.SVR(kernel='linear')\n",
    "train_and_evaluate(clf_svr,X_train,y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf_svr_poly= svm.SVR(kernel='poly')\n",
    "train_and_evaluate(clf_svr_poly,X_train,y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf_svr_rbf= svm.SVR(kernel='rbf')\n",
    "train_and_evaluate(clf_svr_rbf,X_train,y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "clf_svr_poly2= svm.SVR(kernel='poly',degree=2)\n",
    "train_and_evaluate(clf_svr_poly2,X_train,y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally, let's try again Random Forests, in their Extra Trees, and Regression version"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import ensemble\n",
    "clf_et=ensemble.ExtraTreesRegressor(n_estimators=10,random_state=42)\n",
    "train_and_evaluate(clf_et,X_train,y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(clf_et.feature_importances_,boston.feature_names)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally, evaluate our classifiers on the testing set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn import metrics\n",
    "def measure_performance(X,y,clf, show_accuracy=True, show_classification_report=True, show_confusion_matrix=True, show_r2_score=False):\n",
    "    y_pred=clf.predict(X)   \n",
    "    if show_accuracy:\n",
    "        print(\"Accuracy:{0:.3f}\".format(metrics.accuracy_score(y,y_pred)),\"\\n\")\n",
    "\n",
    "    if show_classification_report:\n",
    "        print(\"Classification report\")\n",
    "        print(metrics.classification_report(y,y_pred),\"\\n\")\n",
    "        \n",
    "    if show_confusion_matrix:\n",
    "        print(\"Confusion matrix\")\n",
    "        print(metrics.confusion_matrix(y,y_pred),\"\\n\")\n",
    "        \n",
    "    if show_r2_score:\n",
    "        print(\"Coefficient of determination:{0:.3f}\".format(metrics.r2_score(y,y_pred)),\"\\n\")\n",
    "\n",
    "        \n",
    "measure_performance(X_test,y_test,clf_et, show_accuracy=False, show_classification_report=False,show_confusion_matrix=False, show_r2_score=True)\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "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": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}