NB.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn import discriminant_analysis\n",
    "%matplotlib inline\n",
    "data = pd.read_csv('numeric sequence.csv', sep = ',', engine = 'python')\n",
    "processed_data = data\n",
    "for i in range (1, 15):\n",
    "     processed_data['feature'+str(i)] = data['feature'+str(i*2-1)]*2 + data['feature'+str(i*2)]\n",
    "to_drop=[]\n",
    "for i in range (15, 29):\n",
    "    to_drop.append('feature'+str(i)) \n",
    "processed_data = processed_data.drop(to_drop, 1).drop('label', 1)\n",
    "processed_data = processed_data.replace(0, 'A')\\\n",
    "                              .replace(1, 'T')\\\n",
    "                              .replace(2, 'C')\\\n",
    "                              .replace(3, 'G')\n",
    "processed_data = pd.get_dummies(processed_data)\n",
    "processed_data=pd.concat((processed_data, data['label']), axis = 1)\n",
    "data = processed_data\n",
    "X, X_test, y, y_test = train_test_split(data.drop('label', axis=1), data['label'], test_size = 0.33, \n",
    "                                                    random_state = 2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Наивный байесовский классификатор"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "GaussianNB(priors=None)"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.naive_bayes import GaussianNB\n",
    "nb_model = GaussianNB()\n",
    "nb_model.fit(X, y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "$\n",
    "\\begin{array}{|c|c|}\n",
    "\\hline\n",
    "{\\rm TN} & {\\rm FN} \\\\\n",
    "\\hline\n",
    "{\\rm FP} & {\\rm TP} \\\\\n",
    "\\hline\n",
    "\\end{array}\n",
    "$"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.148631840796 0.152777777778\n",
      "[[705 142]\n",
      " [ 97 664]]\n",
      "[[342  65]\n",
      " [ 56 329]]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import confusion_matrix\n",
    "# Ошибки на обучающей и тестовойвыборке\n",
    "print (np.mean(y != nb_model.predict(X)), np.mean(y_test != nb_model.predict(X_test)))\n",
    "print(confusion_matrix(nb_model.predict(X), y)) # confusion matrix for train sample\n",
    "print(confusion_matrix(nb_model.predict(X_test), y_test)) # confusion matrix for test sample"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.147388059701 0.15404040404\n",
      "[[690 125]\n",
      " [112 681]]\n",
      "[[335  59]\n",
      " [ 63 335]]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.naive_bayes import MultinomialNB\n",
    "nb_model = MultinomialNB() # another model\n",
    "nb_model.fit(X, y)\n",
    "# Ошибки на обучающей и тестовойвыборке\n",
    "print (np.mean(y != nb_model.predict(X)), np.mean(y_test != nb_model.predict(X_test)))\n",
    "print(confusion_matrix(nb_model.predict(X), y)) # confusion matrix for train sample\n",
    "print(confusion_matrix(nb_model.predict(X_test), y_test)) # confusion matrix for test sample"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.14552238806 0.14898989899\n",
      "[[693 125]\n",
      " [109 681]]\n",
      "[[339  59]\n",
      " [ 59 335]]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.naive_bayes import BernoulliNB\n",
    "nb_model = BernoulliNB()\n",
    "nb_model.fit(X, y)\n",
    "# Ошибки на обучающей и тестовойвыборке\n",
    "print (np.mean(y != nb_model.predict(X)), np.mean(y_test != nb_model.predict(X_test)))\n",
    "print(confusion_matrix(nb_model.predict(X), y)) # confusion matrix for train sample\n",
    "print(confusion_matrix(nb_model.predict(X_test), y_test)) # confusion matrix for test sample"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## ROC-кривая"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.metrics import roc_curve\n",
    "proba = nb_model.predict_proba(X_test)[:, 1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAD8CAYAAAB5Pm/hAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x23091578208>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(proba[y_test == 0], color = 'b', alpha = 0.5, normed = False)\n",
    "plt.hist(proba[y_test == 1], color = 'r', alpha = 0.5, normed = False)\n",
    "pass"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[-0.05, 1.05, -0.05, 1.05]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x230951090f0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fpr, tpr, thresholds = roc_curve(y_test, proba)\n",
    "plt.title('Receiver Operating Characteristic')\n",
    "plt.plot(np.concatenate(([0], fpr)), np.concatenate(([0], tpr)))\n",
    "#plt.plot(fpr, tpr)\n",
    "plt.xlabel('False Positive Rate (1 - Specificity)')\n",
    "plt.ylabel('True Positive Rate (Sensitivity)')\n",
    "plt.plot([0, 1], [0, 1], 'k--')\n",
    "plt.axis([-0.05, 1.05, -0.05, 1.05])"
   ]
  },
  {
   "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": 2
}