bagged_scikit_neuralnetwork.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# bagged_scikit_neuralnetwork"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from __future__ import division\n",
    "from IPython.display import display\n",
    "from matplotlib import pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "import numpy  as np\n",
    "import pandas as pd\n",
    "import random, sys, os, re\n",
    "\n",
    "from sknn.mlp                 import Classifier, Layer\n",
    "\n",
    "from sklearn.cross_validation import StratifiedKFold\n",
    "from sklearn.grid_search      import RandomizedSearchCV, GridSearchCV\n",
    "from sklearn.cross_validation import cross_val_predict, permutation_test_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "SEED   = 97\n",
    "scale  = True \n",
    "minmax = False\n",
    "norm   = False\n",
    "nointercept = True\n",
    "engineering = True\n",
    "\n",
    "N_CLASSES  = 2\n",
    "max_epochs = 100\n",
    "\n",
    "submission_filename = \"../submissions/submission_bagged_scikit_nn.csv\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "source": [
    "# Load the training data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from load_blood_data import load_blood_data\n",
    "\n",
    "y_train, X_train = load_blood_data(train=True, SEED   = SEED, \n",
    "                                               scale  = scale,\n",
    "                                               minmax = minmax,\n",
    "                                               norm   = norm,\n",
    "                                               nointercept = nointercept,\n",
    "                                               engineering = engineering)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fit the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "StatifiedCV = StratifiedKFold(y            = y_train, \n",
    "                              n_folds      = 10, \n",
    "                              shuffle      = True, \n",
    "                              random_state = SEED)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "\"\"\"\n",
    "[\"Bagging Predictors\" Department of Statistics UC Berkeley 1994; Leo Breiman]\n",
    "\"\"\"\n",
    "\n",
    "from sklearn.base import BaseEstimator, ClassifierMixin, clone\n",
    "\n",
    "\n",
    "class BaggedScikitClassifier(BaseEstimator, ClassifierMixin):\n",
    "    \"\"\"\n",
    "    Runs a scikit-learn base estimator 'n_estimators' times, fitting with\n",
    "    the X_train indices drawn from a random sample of size X_train.shape[0] \n",
    "    with replacement\n",
    "    \n",
    "    sklearn.ensemble.BaggingClassifier is more convenient and has more tunable\n",
    "    features but it requires the base estimator to have a 'fit' method with a\n",
    "    'sample_weight' parameter and many estimators do not \n",
    "    (as of sklearn.__version__ == '0.17').\n",
    "    \n",
    "    \n",
    "    Usage\n",
    "    -----\n",
    "    clf = BaggedScikitClassifier(base_estimator, n_estimators)\n",
    "    \n",
    "        base_estimator is an instance of a scikit-learn object\n",
    "        n_estimators   is the number of bagged replicas to use\n",
    "        \n",
    "    clf.fit(X_train, y_train)\n",
    "    \n",
    "        a no-op; merely stores X_train and y_train for use in the\n",
    "        predict and predict_proba methods\n",
    "        \n",
    "    clf.predict(X_test)\n",
    "    \n",
    "        the value returned is the majority vote of each of the\n",
    "        fitted base_estimators' class predictions\n",
    "        \n",
    "    clf.predict_proba(X_test)\n",
    "    \n",
    "        the value returned is the average of each of the\n",
    "        fitted base_estimators' probabilities\n",
    "    \"\"\"\n",
    "    def __init__(self, base_estimator, n_estimators=15):\n",
    "        self.n_estimators   = n_estimators\n",
    "        self.base_estimator = base_estimator\n",
    "\n",
    "        \n",
    "    def fit(self, X_train, y_train):\n",
    "        \"\"\"\n",
    "        The actual fitting is done in the predict & predict_proba methods\n",
    "        \"\"\"\n",
    "        self.X_train = X_train\n",
    "        self.y_train = y_train\n",
    "        \n",
    "        return self\n",
    "        \n",
    "            \n",
    "    def predict(self, X_test):\n",
    "        \"\"\"\n",
    "        Return the majority vote of 'n_estimators' fits\n",
    "        \"\"\"\n",
    "        predictions = []\n",
    "        for i in range(self.n_estimators):\n",
    "            clf = clone(self.base_estimator)\n",
    "            idx = np.random.choice(a    = range(self.X_train.shape[0]), \n",
    "                                   size = self.X_train.shape[0], \n",
    "                                   replace=True, p=None)\n",
    "\n",
    "            clf.fit(self.X_train[idx,:], self.y_train[idx])\n",
    "            predictions.append(clf.predict(X_test))\n",
    "            \n",
    "        from scipy.stats import mode\n",
    "        \n",
    "        return mode(predictions)[0][0]\n",
    "    \n",
    "    def predict_proba(self, X_test):\n",
    "        \"\"\"\n",
    "        Return the average probability matrix of 'n_estimators' fits\n",
    "        \"\"\"\n",
    "        completed_bags = 0\n",
    "        predictions = np.empty((X_test.shape[0], N_CLASSES), dtype=np.float32)\n",
    "        for i in range(self.n_estimators):\n",
    "            clf = clone(self.base_estimator)\n",
    "            idx = np.random.choice(a    = range(self.X_train.shape[0]), \n",
    "                                   size = self.X_train.shape[0], \n",
    "                                   replace=True, p=None)\n",
    "            \n",
    "            clf.fit(self.X_train[idx,:], self.y_train[idx])\n",
    "            predictions += clf.predict_proba(X_test)\n",
    "            completed_bags += 1\n",
    "            \n",
    "        return predictions / float(completed_bags)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 0 ns, sys: 0 ns, total: 0 ns\n",
      "Wall time: 133 µs\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "random.seed(SEED)\n",
    "\n",
    "# ----------------------------------- base estimator -------------------------------\n",
    "nn_layers  = [\n",
    "              Layer(type = 'Rectifier', name         = 'hidden', \n",
    "                                        units        = 100,\n",
    "                                        weight_decay = None,\n",
    "                                        pieces       = None,\n",
    "                                        dropout      = None),\n",
    "    \n",
    "              Layer(type = 'Softmax',   name         = 'output')\n",
    "             ]\n",
    "\n",
    "base = Classifier(layers            = nn_layers,\n",
    "                 learning_rate     = 0.01,\n",
    "                 learning_rule     = 'nesterov',\n",
    "                 learning_momentum = 0.9,\n",
    "                 loss_type         = u'mse',\n",
    "                 mutator           = None,    # data augmentation function\n",
    "                 \n",
    "                 regularize        = None,\n",
    "                 weight_decay      = None,\n",
    "                 dropout_rate      = None,\n",
    "                 \n",
    "                 batch_size        = 10,\n",
    "                 \n",
    "                 valid_size        = None,    \n",
    "                 valid_set         = None,\n",
    "                 \n",
    "                 n_stable          = 10,      # early stopping after ...\n",
    "                 f_stable          = 0.001,   # validation error change threshold\n",
    "                 n_iter            = max_epochs,     \n",
    "                 \n",
    "                 random_state      = SEED, \n",
    "                 debug             = False, \n",
    "                 verbose           = True) \n",
    "\n",
    "# ----------------------------------- BaggedScikitClassifier ------------------------------- \n",
    "clf = BaggedScikitClassifier(base_estimator = base, \n",
    "                             n_estimators   = 15)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# from sklearn_utilities import GridSearchHeatmap\n",
    "\n",
    "# GridSearchHeatmap(grid_clf, y_key='learning_rate', x_key='n_estimators')\n",
    "\n",
    "# from sklearn_utilities import plot_validation_curves\n",
    "\n",
    "# plot_validation_curves(grid_clf, param_grid, X_train, y_train, ylim = (0.0, 1.05))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAmYAAAGJCAYAAAAg1v9AAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nIqIhUphFRERENEQKs4iIiIiG+B/r9Y0l7m0qFQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f8c90e67e10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 468 ms, sys: 104 ms, total: 572 ms\n",
      "Wall time: 5min 43s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "try:\n",
    "    from sklearn_utilities import plot_learning_curve\n",
    "except:\n",
    "    import imp, os\n",
    "    util = imp.load_source('sklearn_utilities', os.path.expanduser('~/Dropbox/Python/sklearn_utilities.py'))\n",
    "    from sklearn_utilities import plot_learning_curve\n",
    "\n",
    "plot_learning_curve(estimator   = clf, \n",
    "                    title       = 'Learning Curves', \n",
    "                    X           = X_train.values.astype(np.float32), \n",
    "                    y           = y_train, \n",
    "                    ylim        = (0.0, 1.10), \n",
    "                    cv          = StatifiedCV, \n",
    "                    train_sizes = np.linspace(.1, 1.0, 5),\n",
    "                    n_jobs      = -1)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Training set predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 100 ms, sys: 40 ms, total: 140 ms\n",
      "Wall time: 1min\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "train_preds = cross_val_predict(estimator    = clf, \n",
    "                                X            = X_train.values.astype(np.float32), \n",
    "                                y            = y_train, \n",
    "                                cv           = StatifiedCV, \n",
    "                                n_jobs       = -1, \n",
    "                                verbose      = 0, \n",
    "                                fit_params   = None, \n",
    "                                pre_dispatch = '2*n_jobs')\n",
    "\n",
    "y_true, y_pred   = y_train, train_preds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[400  38]\n",
      " [ 78  60]]\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfsAAAHECAYAAAA+p4OqAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f8c90e79b90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy 0.7986, mis-class rate 0.2014\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import confusion_matrix\n",
    "cm = confusion_matrix(y_true, y_pred, labels=None)\n",
    "print cm\n",
    "\n",
    "try:\n",
    "    from sklearn_utilities import plot_confusion_matrix\n",
    "except:\n",
    "    import imp, os\n",
    "    util = imp.load_source('sklearn_utilities', os.path.expanduser('~/Dropbox/Python/sklearn_utilities.py'))\n",
    "    from sklearn_utilities import plot_confusion_matrix\n",
    "\n",
    "plot_confusion_matrix(cm, ['Did not Donate','Donated'])\n",
    "\n",
    "accuracy = round(np.trace(cm)/float(np.sum(cm)),4)\n",
    "misclass = 1 - accuracy\n",
    "print(\"Accuracy {}, mis-class rate {}\".format(accuracy,misclass))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAmUAAAGJCAYAAADL4URDAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAPwfiw6p/NwQgwkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f8c90e79990>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.metrics import roc_curve\n",
    "from sklearn.metrics import roc_auc_score\n",
    "from sklearn.metrics import log_loss\n",
    "from sklearn.metrics import f1_score\n",
    "\n",
    "fpr, tpr, thresholds = roc_curve(y_true, y_pred, pos_label=None)\n",
    "\n",
    "\n",
    "plt.figure(figsize=(10,6))\n",
    "plt.plot([0, 1], [0, 1], 'k--')\n",
    "plt.plot(fpr, tpr)\n",
    "\n",
    "AUC = roc_auc_score(y_true, y_pred, average='macro')\n",
    "plt.text(x=0.6,y=0.4,s=\"AUC         {:.4f}\"\\\n",
    "         .format(AUC),\n",
    "        fontsize=16)\n",
    "\n",
    "plt.text(x=0.6,y=0.3,s=\"accuracy {:.2f}%\"\\\n",
    "         .format(accuracy*100),\n",
    "        fontsize=16)\n",
    "\n",
    "logloss = log_loss(y_true, y_pred)\n",
    "plt.text(x=0.6,y=0.2,s=\"LogLoss   {:.4f}\"\\\n",
    "         .format(logloss),\n",
    "        fontsize=16)\n",
    "\n",
    "f1 = f1_score(y_true, y_pred)\n",
    "plt.text(x=0.6,y=0.1,s=\"f1             {:.4f}\"\\\n",
    "         .format(f1),\n",
    "        fontsize=16)\n",
    "\n",
    "plt.xlabel('False positive rate')\n",
    "plt.ylabel('True positive rate')\n",
    "plt.title('ROC curve')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABIMAAAH9CAYAAACJG7zhAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAIAO+f8EpZCMdm5aGwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f8c8d4a8e90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 3min 10s, sys: 1.33 s, total: 3min 12s\n",
      "Wall time: 1h 23min 30s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "# score : The true score without permuting targets.\n",
    "# permutation_scores : array, shape (n_permutations,)\n",
    "#                     The scores obtained for each permutation.\n",
    "# pvalue : The returned value equals \n",
    "#              > p-value if scoring returns bigger numbers for better scores (e.g., accuracy_score). \n",
    "#              > If scoring is a loss function (i.e. when lower is better such as with mean_squared_error) \n",
    "#                then this is actually the complement of the p-value: 1 - p-value.\n",
    "\n",
    "score, permutation_scores, pvalue = permutation_test_score(estimator      = clf, \n",
    "                                                           X              = X_train.values.astype(np.float32), \n",
    "                                                           y              = y_train, \n",
    "                                                           cv             = StatifiedCV, \n",
    "                                                           labels         = None,\n",
    "                                                           random_state   = SEED,\n",
    "                                                           verbose        = 0,\n",
    "                                                           n_permutations = 100, \n",
    "                                                           scoring        = None,\n",
    "                                                           n_jobs         = -1) \n",
    "# find mean and stdev of the scores\n",
    "from scipy.stats import norm\n",
    "mu, std = norm.fit(permutation_scores)\n",
    "\n",
    "\n",
    "plt.figure(figsize=(20,8))\n",
    "plt.hist(permutation_scores, 20, label='Permutation scores')\n",
    "\n",
    "ylim = plt.ylim()\n",
    "plt.plot(2 * [score], ylim, '--g', linewidth=3,\n",
    "         label='Classification Score (pvalue {:.4f})'.format(pvalue))\n",
    "         \n",
    "plt.plot(2 * [1. / N_CLASSES], ylim, 'r', linewidth=7, label='Luck')\n",
    "\n",
    "plt.ylim(ylim)\n",
    "plt.title('mu={:.4f}, std={:.4f}'.format(mu,std), fontsize=20)\n",
    "plt.legend(loc='center',fontsize=16)\n",
    "plt.xlabel('Score')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "bagged_scikit_nn           ,        ,   0.7986 ,  6.9558 , 0.6740 , 0.5085 , 0.7463 , 0.0065\n"
     ]
    }
   ],
   "source": [
    "# format for scores.csv file\n",
    "import re\n",
    "algo = re.search(r\"submission_(.*?)\\.csv\", submission_filename).group(1)\n",
    "print(\"{: <26} ,        ,   {:.4f} ,  {:.4f} , {:.4f} , {:.4f} , {:.4f} , {:.4f}\"\\\n",
    "      .format(algo,accuracy,logloss,AUC,f1,mu,std))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Predict the leaderboard score with linear regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "clear all but functions rm(list = setdiff(ls(), lsf.str()))\n",
      "R functions in Dropbox:\n",
      " [1] \"confusion.matrix_ROC.R\" \"eda.R\"                  \"knitr_settings.R\"      \n",
      " [4] \"lm_utilities.R\"         \"lsp.R\"                  \"matrix.power.R\"        \n",
      " [7] \"metrics.R\"              \"multiplot.R\"            \"my.special.functions.R\"\n",
      "[10] \"NOT_IN.R\"               \"SDAFE.R\"                \"signOnToTwitter.R\"     \n",
      "[13] \"source.dropbox.R\"       \"stem_plot.R\"           \n",
      "to load these functions: source(\"/home/george/Dropbox/R_functions/*.R\")\n",
      "\n",
      "To save graphical parameters: opar = par(no.readonly=TRUE)\n",
      "to restore graphical parameters: par(opar)\n"
     ]
    }
   ],
   "source": [
    "# load the R extension\n",
    "%load_ext rpy2.ipython\n",
    "\n",
    "# see http://ipython.readthedocs.org/en/stable/config/extensions/index.html?highlight=rmagic\n",
    "# see http://rpy.sourceforge.net/rpy2/doc-2.4/html/interactive.html#module-rpy2.ipython.rmagic"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Import python variables into R\n",
    "%R -i accuracy,logloss,AUC,f1,mu,std"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "     fit    lwr    upr\n",
       "1 0.8084 0.0429 1.5738\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%%R\n",
    "# read in the scores.csv file and perform a linear regression with it using this process's variables\n",
    "\n",
    "score_data = read.csv('../input/scores.csv')\n",
    "\n",
    "lm.fit     = lm(leaderboard_score ~ accuracy + logloss + AUC + f1 + mu + std, \n",
    "                data      = score_data, \n",
    "                na.action = na.omit)\n",
    "\n",
    "slm.fit = step(lm.fit, direction = \"both\", trace=0)\n",
    "\n",
    "predicted_leaderboard_score = predict(object   = slm.fit, \n",
    "                                      newdata  = data.frame(accuracy,logloss,AUC,f1,mu,std),\n",
    "                                      interval = \"prediction\", level = 0.99)\n",
    "\n",
    "print(round(predicted_leaderboard_score,4))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# --------------------------------------------------------------------------------------------"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# Test Set Predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Re-fit with the full training set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "BaggedScikitClassifier(base_estimator=Classifier(batch_size=10, debug=False, dropout_rate=None, f_stable=0.001,\n",
       "      hidden=<sknn.nn.Layer `Rectifier`: name='hidden', units=100>,\n",
       "      layers=[<sknn.nn.Layer `Rectifier`: name='hidden', units=100>, <sknn.nn.Layer `Softmax`: name='output'>],\n",
       "      learning_momentum=0.9, le...e=97,\n",
       "      regularize=None, valid_set=None, valid_size=None, verbose=True,\n",
       "      weight_decay=None),\n",
       "            n_estimators=15)"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#clf.set_params(**clf_params)\n",
    "clf.fit(X_train.values.astype(np.float32), y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Load the test data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from load_blood_data import load_blood_data\n",
    "\n",
    "X_test, IDs = load_blood_data(train=False, SEED   = SEED, \n",
    "                                           scale  = scale,\n",
    "                                           minmax = minmax,\n",
    "                                           norm   = norm,\n",
    "                                           nointercept = nointercept,\n",
    "                                           engineering = engineering)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Predict the test set with the fitted model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1 0 0 0 1 1 0 0 0 0]\n",
      "[[  5.32215059e-01   4.67784971e-01]\n",
      " [  8.90477240e-01   1.09522872e-01]\n",
      " [  8.65536153e-01   1.34463772e-01]\n",
      " [  8.86783600e-01   1.13216370e-01]\n",
      " [  4.22802091e-01   5.77197909e-01]\n",
      " [  1.89969927e-01   8.10029984e-01]\n",
      " [  7.37736285e-01   2.62263745e-01]\n",
      " [  9.58714843e-01   4.12850343e-02]\n",
      " [  9.99967933e-01   3.19885330e-05]\n",
      " [  9.84968185e-01   1.50317950e-02]]\n",
      "[0.46778497, 0.10952287, 0.13446377, 0.11321637, 0.57719791, 0.81002998, 0.26226375, 0.041285034, 3.1988533e-05, 0.015031795]\n"
     ]
    }
   ],
   "source": [
    "y_pred = clf.predict(X_test.values.astype(np.float32)).ravel()\n",
    "print(y_pred[:10])\n",
    "\n",
    "try:\n",
    "    y_pred_probs  = clf.predict_proba(X_test.values.astype(np.float32))\n",
    "    print(y_pred_probs[:10])\n",
    "    donate_probs  = [prob[1] for prob in y_pred_probs]\n",
    "except Exception,e:\n",
    "    print(e)\n",
    "    donate_probs = [0.65 if x>0 else 1-0.65 for x in y_pred]\n",
    "    \n",
    "print(donate_probs[:10])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Create the submission file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "assert len(IDs)==len(donate_probs)\n",
    "\n",
    "f = open(submission_filename, \"w\")\n",
    "\n",
    "f.write(\",Made Donation in March 2007\\n\")\n",
    "for ID, prob in zip(IDs, donate_probs):\n",
    "    f.write(\"{},{}\\n\".format(ID,prob))\n",
    "    \n",
    "f.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}