09_eval-ci/09_eval-ci_code.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "STAT 479: Machine Learning (Fall 2018)  \n",
    "Instructor: Sebastian Raschka (sraschka@wisc.edu)  \n",
    "Course website: http://pages.stat.wisc.edu/~sraschka/teaching/stat479-fs2018/"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# L09: Model Evaluation 2 -- Confidence Intervals and Bootstrapping"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sebastian Raschka \n",
      "last updated: 2018-10-24 \n",
      "\n",
      "CPython 3.6.6\n",
      "IPython 6.5.0\n",
      "\n",
      "numpy 1.15.1\n",
      "mlxtend 0.14.0dev\n",
      "matplotlib 2.2.3\n",
      "sklearn 0.20.0\n"
     ]
    }
   ],
   "source": [
    "%load_ext watermark\n",
    "%watermark -a 'Sebastian Raschka' -u -d -v -p numpy,mlxtend,matplotlib,sklearn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Iris Feature by Class Distribution in Random Subsampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from mlxtend.data import iris_data\n",
    "from mlxtend.preprocessing import shuffle_arrays_unison\n",
    "\n",
    "X, y = iris_data()\n",
    "\n",
    "X, y = shuffle_arrays_unison([X, y], random_seed=123)\n",
    "X_train, X_test = X[:100], X[100:150]\n",
    "y_train, y_test = y[:100], y[100:150]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([50, 50, 50])"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.bincount(y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([38, 28, 34])"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.bincount(y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([12, 22, 16])"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.bincount(y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.DataFrame(X)\n",
    "df['class'] = y\n",
    "\n",
    "df_train = pd.DataFrame(X_train)\n",
    "df_train['class'] = y_train\n",
    "\n",
    "df_test = pd.DataFrame(X_test)\n",
    "df_test['class'] = y_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def stackhist(x, y, **kws):\n",
    "    grouped = x.groupby(y)\n",
    "    data = [d for _, d in grouped]\n",
    "    labels = [l for l, _ in grouped]\n",
    "    plt.hist(data, \n",
    "             histtype=\"barstacked\", \n",
    "             label=labels, \n",
    "             alpha=0.8, \n",
    "             density=True,\n",
    "             bins=np.arange(4.0, 8.1, 0.25))\n",
    "    plt.ylim([0, 0.8])\n",
    "    plt.xlim([4, 8])\n",
    "    plt.xlabel('Sepal Width [cm]')\n",
    "    plt.ylabel('Frequency')\n",
    "    \n",
    "\n",
    "with plt.style.context('fivethirtyeight'):    \n",
    "\n",
    "    stackhist(df[0], df['class'])\n",
    "    plt.legend(['Setosa', 'Virginica', 'Versicolor'], fontsize=12)\n",
    "    plt.title('All')\n",
    "    plt.tight_layout()\n",
    "    #plt.savefig('./all.svg')\n",
    "    plt.show()    \n",
    "\n",
    "    stackhist(df_train[0], df_train['class'])\n",
    "    plt.title('Train')\n",
    "    plt.tight_layout()\n",
    "    #plt.savefig('./train.svg')\n",
    "    plt.show()    \n",
    "\n",
    "    stackhist(df_test[0], df_test['class'])\n",
    "    plt.title('Test')\n",
    "    plt.tight_layout()\n",
    "    #plt.savefig('./test.svg')\n",
    "    plt.show() \n",
    "    "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Holdout method and repeated sampling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "from mlxtend.data import iris_data\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "\n",
    "X, y = iris_data()\n",
    "\n",
    "clf_1 = KNeighborsClassifier(n_neighbors=3,\n",
    "                             weights='uniform', \n",
    "                             algorithm='kd_tree', \n",
    "                             leaf_size=30, \n",
    "                             p=2, \n",
    "                             metric='minkowski', \n",
    "                             metric_params=None, \n",
    "                             n_jobs=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average: 95.41%\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "rng = np.random.RandomState(seed=12345)\n",
    "seeds = np.arange(10**5)\n",
    "rng.shuffle(seeds)\n",
    "seeds = seeds[:50]\n",
    "\n",
    "pred_2 = []\n",
    "\n",
    "for i in seeds:\n",
    "    X_train, X_test, y_train, y_test = train_test_split(X, y,\n",
    "                                                        test_size=0.5, \n",
    "                                                        random_state=i,\n",
    "                                                        stratify=y)\n",
    "    y_pred_i = clf_1.fit(X_train, y_train).predict(X_test)\n",
    "    y_pred_i_acc = np.mean(y_test == y_pred_i)\n",
    "    pred_2.append(y_pred_i_acc)\n",
    "\n",
    "pred_2 = np.asarray(pred_2)\n",
    "print('Average: %.2f%%' % (pred_2.mean()*100))\n",
    "\n",
    "with plt.style.context(('fivethirtyeight')):\n",
    "    plt.bar(range(0, pred_2.shape[0]), pred_2, color='gray', alpha=0.7)\n",
    "    plt.axhline(pred_2.max(), color='k', linewidth=1, linestyle='--')\n",
    "    plt.axhline(pred_2.min(), color='k', linewidth=1, linestyle='--')\n",
    "    plt.axhspan(pred_2.min(), pred_2.max(), alpha=0.2, color='steelblue')\n",
    "    plt.ylim([0, pred_2.max() + 0.1])\n",
    "    plt.xlabel('Repetition')\n",
    "    plt.ylabel('Accuracy')\n",
    "    plt.ylim([0.5, 1.05])\n",
    "    plt.tight_layout()\n",
    "    #plt.savefig('figures/model-eval-iris_0.svg')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Average: 96.13%\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "rng = np.random.RandomState(seed=12345)\n",
    "seeds = np.arange(10**5)\n",
    "rng.shuffle(seeds)\n",
    "seeds = seeds[:50]\n",
    "\n",
    "pred_2 = []\n",
    "\n",
    "for i in seeds:\n",
    "    X_train, X_test, y_train, y_test = train_test_split(X, y,\n",
    "                                                        test_size=0.1, \n",
    "                                                        random_state=i,\n",
    "                                                        stratify=y)\n",
    "    y_pred_i = clf_1.fit(X_train, y_train).predict(X_test)\n",
    "    y_pred_i_acc = np.mean(y_test == y_pred_i)\n",
    "    pred_2.append(y_pred_i_acc)\n",
    "    \n",
    "pred_2 = np.asarray(pred_2)\n",
    "print('Average: %.2f%%' % (pred_2.mean()*100))\n",
    "\n",
    "with plt.style.context(('fivethirtyeight')):\n",
    "    plt.bar(range(0, pred_2.shape[0]), pred_2, color='gray', alpha=0.7)\n",
    "    plt.axhline(pred_2.max(), color='k', linewidth=1, linestyle='--')\n",
    "    plt.axhline(pred_2.min(), color='k', linewidth=1, linestyle='--')\n",
    "    plt.axhspan(pred_2.min(), pred_2.max(), alpha=0.2, color='steelblue')\n",
    "    plt.ylim([0, pred_2.max() + 0.1])\n",
    "    plt.xlabel('Repetition')\n",
    "    plt.ylabel('Accuracy')\n",
    "    plt.ylim(0.5, 1.05)\n",
    "    plt.tight_layout()\n",
    "    #plt.savefig('figures/model-eval-iris_0_2.svg')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Pessimistic Bias in Holdout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of train examples: 3500\n",
      "Number of test examples: 1500\n",
      "Labels: [8 8 7 ... 6 9 8]\n"
     ]
    }
   ],
   "source": [
    "from mlxtend.data import mnist_data\n",
    "\n",
    "X2, y2 = mnist_data()\n",
    "X_train2, X_test2, y_train2, y_test2 = train_test_split(X2, y2,\n",
    "                                                        test_size=0.3, \n",
    "                                                        random_state=12,\n",
    "                                                        stratify=y2)\n",
    "print('Number of train examples:', y_train2.shape[0])\n",
    "print('Number of test examples:', y_test2.shape[0])\n",
    "print('Labels:', y_train2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "\n",
    "clf_2 = LogisticRegression(penalty='l2', \n",
    "                           dual=False, \n",
    "                           tol=0.0001, \n",
    "                           C=0.000001, \n",
    "                           fit_intercept=True, \n",
    "                           intercept_scaling=1, \n",
    "                           class_weight=None, \n",
    "                           random_state=12, \n",
    "                           solver='lbfgs', \n",
    "                           max_iter=1000, \n",
    "                           multi_class='multinomial', \n",
    "                           verbose=0, \n",
    "                           warm_start=False, \n",
    "                           n_jobs=1)\n",
    "\n",
    "pred_train, pred_test = [], []\n",
    "\n",
    "intervals = np.arange(500, X_train2.shape[0] + 1, 200)\n",
    "\n",
    "for i in intervals:\n",
    "    clf_2.fit(X_train2[:i], y_train2[:i])\n",
    "    p_train = clf_2.score(X_train2[:i], y_train2[:i])\n",
    "    p_test = clf_2.score(X_test2, y_test2)\n",
    "    pred_train.append(p_train)\n",
    "    pred_test.append(p_test)\n",
    "    \n",
    "    \n",
    "with plt.style.context(('fivethirtyeight')):\n",
    "    plt.plot(intervals, pred_train, marker='o', label='Train')\n",
    "    plt.plot(intervals, pred_test, marker='s', label='Test')\n",
    "    plt.legend(loc='best', numpoints=1)\n",
    "    plt.xlim([430, X_train2.shape[0] + X_test2.shape[0]])\n",
    "    plt.axvspan(X_train2.shape[0], \n",
    "                X_train2.shape[0] + X_test2.shape[0], \n",
    "                alpha=0.2, \n",
    "                color='steelblue')\n",
    "    plt.ylim([0.85, 1.0])\n",
    "    plt.xlabel('Training Set Size')\n",
    "    plt.ylabel('Accuracy')\n",
    "    plt.tight_layout()\n",
    "    #plt.savefig('figures/model-eval-mnist_0.svg')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier\n",
    "\n",
    "\n",
    "clf_2 = RandomForestClassifier(n_estimators=100, random_state=123)\n",
    "\n",
    "pred_train, pred_test = [], []\n",
    "\n",
    "intervals = np.arange(500, X_train2.shape[0] + 1, 200)\n",
    "\n",
    "for i in intervals:\n",
    "    clf_2.fit(X_train2[:i], y_train2[:i])\n",
    "    p_train = clf_2.score(X_train2[:i], y_train2[:i])\n",
    "    p_test = clf_2.score(X_test2, y_test2)\n",
    "    pred_train.append(p_train)\n",
    "    pred_test.append(p_test)\n",
    "    \n",
    "    \n",
    "with plt.style.context(('fivethirtyeight')):\n",
    "    plt.plot(intervals, pred_train, marker='o', label='Train')\n",
    "    plt.plot(intervals, pred_test, marker='s', label='Test')\n",
    "    plt.legend(loc='best', numpoints=1)\n",
    "    plt.xlim([430, X_train2.shape[0] + X_test2.shape[0]])\n",
    "    plt.axvspan(X_train2.shape[0], \n",
    "                X_train2.shape[0] + X_test2.shape[0], \n",
    "                alpha=0.2, \n",
    "                color='steelblue')\n",
    "    plt.ylim([0.85, 1.0])\n",
    "    plt.xlabel('Training Set Size')\n",
    "    plt.ylabel('Accuracy')\n",
    "    plt.tight_layout()\n",
    "    #plt.savefig('figures/model-eval-mnist_0.svg')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "\n",
    "\n",
    "clf_2 = KNeighborsClassifier(n_neighbors=3)\n",
    "\n",
    "pred_train, pred_test = [], []\n",
    "\n",
    "intervals = np.arange(500, X_train2.shape[0] + 1, 200)\n",
    "\n",
    "for i in intervals:\n",
    "    clf_2.fit(X_train2[:i], y_train2[:i])\n",
    "    p_train = clf_2.score(X_train2[:i], y_train2[:i])\n",
    "    p_test = clf_2.score(X_test2, y_test2)\n",
    "    pred_train.append(p_train)\n",
    "    pred_test.append(p_test)\n",
    "    \n",
    "    \n",
    "with plt.style.context(('fivethirtyeight')):\n",
    "    plt.plot(intervals, pred_train, marker='o', label='Train')\n",
    "    plt.plot(intervals, pred_test, marker='s', label='Test')\n",
    "    plt.legend(loc='best', numpoints=1)\n",
    "    plt.xlim([430, X_train2.shape[0] + X_test2.shape[0]])\n",
    "    plt.axvspan(X_train2.shape[0], \n",
    "                X_train2.shape[0] + X_test2.shape[0], \n",
    "                alpha=0.2, \n",
    "                color='steelblue')\n",
    "    plt.ylim([0.85, 1.0])\n",
    "    plt.xlabel('Training Set Size')\n",
    "    plt.ylabel('Accuracy')\n",
    "    plt.tight_layout()\n",
    "    #plt.savefig('figures/model-eval-mnist_0.svg')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##  Bootstrapping"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Iris"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "\n",
    "rng = np.random.RandomState(seed=12345)\n",
    "\n",
    "X, y = iris_data()\n",
    "\n",
    "clf = KNeighborsClassifier(n_neighbors=3,\n",
    "                             weights='uniform', \n",
    "                             algorithm='kd_tree', \n",
    "                             leaf_size=30, \n",
    "                             p=2, \n",
    "                             metric='minkowski', \n",
    "                             metric_params=None, \n",
    "                             n_jobs=1)\n",
    "\n",
    "\n",
    "idx = np.arange(y.shape[0])\n",
    "\n",
    "accuracies = []\n",
    "\n",
    "for i in range(200):\n",
    "    \n",
    "    train_idx = rng.choice(idx, size=idx.shape[0], replace=True)\n",
    "    test_idx = np.setdiff1d(idx, train_idx, assume_unique=False)\n",
    "    \n",
    "    boot_train_X, boot_train_y = X[train_idx], y[train_idx]\n",
    "    boot_test_X, boot_test_y = X[test_idx], y[test_idx]\n",
    "    \n",
    "    clf.fit(boot_train_X, boot_train_y)\n",
    "    acc = clf.score(boot_test_X, boot_test_y)\n",
    "    accuracies.append(acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "mean = np.mean(accuracies)\n",
    "\n",
    "#se = np.sqrt( (1. / (100-1)) * np.sum([(acc - mean)**2 for acc in accuracies])) \n",
    "#ci = 1.984 * se\n",
    "\n",
    "se = np.sqrt( (1. / (200-1)) * np.sum([(acc - mean)**2 for acc in accuracies])) \n",
    "ci = 1.97 * se\n",
    "\n",
    "lower = np.percentile(accuracies, 2.5)\n",
    "upper = np.percentile(accuracies, 97.5)\n",
    "\n",
    "fig, ax = plt.subplots(figsize=(8, 4))\n",
    "ax.vlines(mean, [0], 80, lw=2.5, linestyle='-', label='mean')\n",
    "#ax.vlines(med, [0], 60, lw=2.5, linestyle='--', label='median')\n",
    "ax.vlines(lower, [0], 15, lw=2.5, linestyle='-.', label='CI95 percentile')\n",
    "ax.vlines(upper, [0], 15, lw=2.5, linestyle='-.')\n",
    "\n",
    "ax.vlines(mean + ci, [0], 15, lw=2.5, linestyle=':', label='CI95 standard')\n",
    "ax.vlines(mean - ci, [0], 15, lw=2.5, linestyle=':')\n",
    "\n",
    "\n",
    "ax.hist(accuracies, bins=7,\n",
    "        color='#0080ff', edgecolor=\"none\", \n",
    "        alpha=0.3)\n",
    "plt.legend(loc='upper left')\n",
    "\n",
    "plt.xlim([0.86, 1.01])\n",
    "plt.tight_layout()\n",
    "#plt.savefig('figures/bootstrap-histo-1.svg')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5k MNIST Subset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "from mlxtend.data import mnist_data\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "\n",
    "\n",
    "rng = np.random.RandomState(seed=12345)\n",
    "\n",
    "X, y = mnist_data()\n",
    "\n",
    "clf = LogisticRegression(penalty='l2', \n",
    "                           dual=False, \n",
    "                           tol=0.0001, \n",
    "                           C=0.000001, \n",
    "                           fit_intercept=True, \n",
    "                           intercept_scaling=1, \n",
    "                           class_weight=None, \n",
    "                           random_state=12, \n",
    "                           solver='lbfgs', \n",
    "                           max_iter=1000, \n",
    "                           multi_class='multinomial', \n",
    "                           verbose=0, \n",
    "                           warm_start=False, \n",
    "                           n_jobs=1)\n",
    "\n",
    "\n",
    "idx = np.arange(y.shape[0])\n",
    "\n",
    "accuracies = []\n",
    "\n",
    "for i in range(200):\n",
    "    \n",
    "    train_idx = rng.choice(idx, size=idx.shape[0], replace=True)\n",
    "    test_idx = np.setdiff1d(idx, train_idx, assume_unique=False)\n",
    "    \n",
    "    boot_train_X, boot_train_y = X[train_idx], y[train_idx]\n",
    "    boot_test_X, boot_test_y = X[test_idx], y[test_idx]\n",
    "    \n",
    "    clf.fit(boot_train_X, boot_train_y)\n",
    "    acc = clf.score(boot_test_X, boot_test_y)\n",
    "    accuracies.append(acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "mean = np.mean(accuracies)\n",
    "\n",
    "#se = np.sqrt( (1. / (100-1)) * np.sum([(acc - mean)**2 for acc in accuracies])) \n",
    "#ci = 1.984 * se\n",
    "\n",
    "se = np.sqrt( (1. / (200-1)) * np.sum([(acc - mean)**2 for acc in accuracies])) \n",
    "ci = 1.97 * se\n",
    "\n",
    "lower = np.percentile(accuracies, 2.5)\n",
    "upper = np.percentile(accuracies, 97.5)\n",
    "\n",
    "fig, ax = plt.subplots(figsize=(8, 4))\n",
    "ax.vlines(mean, [0], 40, lw=2.5, linestyle='-', label='mean')\n",
    "#ax.vlines(med, [0], 60, lw=2.5, linestyle='--', label='median')\n",
    "ax.vlines(lower, [0], 15, lw=2.5, linestyle='-.', label='CI95 percentile')\n",
    "ax.vlines(upper, [0], 15, lw=2.5, linestyle='-.')\n",
    "\n",
    "ax.vlines(mean + ci, [0], 15, lw=2.5, linestyle=':', label='CI95 standard')\n",
    "ax.vlines(mean - ci, [0], 15, lw=2.5, linestyle=':')\n",
    "\n",
    "\n",
    "ax.hist(accuracies, bins=11,\n",
    "        color='#0080ff', edgecolor=\"none\", \n",
    "        alpha=0.3)\n",
    "plt.legend(loc='upper left')\n",
    "\n",
    "plt.xlim([0.885, 0.922])\n",
    "plt.tight_layout()\n",
    "\n",
    "# plt.savefig('figures/bootstrap-histo-2.svg')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Out of Bag Bootstrap"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1 1 4 1 1] [0 2 3]\n",
      "[0 1 1 3 0] [2 4]\n",
      "[3 3 3 3 0] [1 2 4]\n"
     ]
    }
   ],
   "source": [
    "from mlxtend.evaluate import BootstrapOutOfBag\n",
    "import numpy as np\n",
    "\n",
    "\n",
    "oob = BootstrapOutOfBag(n_splits=3)\n",
    "for train, test in oob.split(np.array([1, 2, 3, 4, 5])):\n",
    "    print(train, test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "from mlxtend.data import iris_data\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.model_selection import cross_val_score\n",
    "\n",
    "X, y = iris_data()\n",
    "model = DecisionTreeClassifier()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CV scores [0.96666667 0.96666667 0.9        1.         1.        ]\n",
      "Mean CV score 0.9666666666666668\n",
      "CV score Std 0.036514837167011066\n"
     ]
    }
   ],
   "source": [
    "cv_scores = cross_val_score(model, X, y, cv=5)\n",
    "print('CV scores', cv_scores)\n",
    "print('Mean CV score', np.mean(cv_scores))\n",
    "print('CV score Std', np.std(cv_scores))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Bootstrap scores [0.92727273 0.98113208 0.92592593 0.98305085 0.90909091]\n",
      "Mean Bootstrap score 0.9452944970437775\n",
      "Score Std 0.030726208804992074\n"
     ]
    }
   ],
   "source": [
    "bootstrap_scores = \\\n",
    "    cross_val_score(model, X, y, \n",
    "                    cv=BootstrapOutOfBag(n_splits=5, random_seed=456))\n",
    "\n",
    "print('Bootstrap scores', bootstrap_scores)\n",
    "print('Mean Bootstrap score', np.mean(bootstrap_scores))\n",
    "print('Score Std', np.std(bootstrap_scores))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean Bootstrap score 0.9448327329412445\n",
      "Score Std 0.030836791007487842\n"
     ]
    }
   ],
   "source": [
    "bootstrap_scores = \\\n",
    "    cross_val_score(model, X, y, \n",
    "                    cv=BootstrapOutOfBag(n_splits=200, random_seed=456))\n",
    "\n",
    "print('Mean Bootstrap score', np.mean(bootstrap_scores))\n",
    "print('Score Std', np.std(bootstrap_scores))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "95% Confidence interval: [87.27, 100.00]\n"
     ]
    }
   ],
   "source": [
    "lower = np.percentile(bootstrap_scores, 2.5)\n",
    "upper = np.percentile(bootstrap_scores, 97.5)\n",
    "print('95%% Confidence interval: [%.2f, %.2f]' % (100*lower, 100*upper))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###  .632 Bootstrap"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean Bootstrap score 0.9627083103060582\n",
      "Score Std 0.016892377745312225\n"
     ]
    }
   ],
   "source": [
    "from mlxtend.evaluate import bootstrap_point632_score\n",
    "\n",
    "\n",
    "bootstrap_scores = bootstrap_point632_score(model, X, y, n_splits=200)\n",
    "print('Mean Bootstrap score', np.mean(bootstrap_scores))\n",
    "print('Score Std', np.std(bootstrap_scores))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "95% Confidence interval: [92.81, 98.89]\n"
     ]
    }
   ],
   "source": [
    "lower = np.percentile(bootstrap_scores, 2.5)\n",
    "upper = np.percentile(bootstrap_scores, 97.5)\n",
    "print('95%% Confidence interval: [%.2f, %.2f]' % (100*lower, 100*upper))"
   ]
  }
 ],
 "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": "3.6.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}