Ch4 Classification.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Classification\n",
    ">* Logistic Regression\n",
    ">* LDA\n",
    ">* QDA\n",
    ">* KNN"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Lab for Logistic Regression, LDA, QDA and KNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "#load module\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "import statsmodels.api as sm\n",
    "from sklearn import metrics\n",
    "from sklearn.cross_validation import train_test_split\n",
    "import matplotlib as mpl\n",
    "from sklearn.metrics import classification_report"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 1250 entries, 0 to 1249\n",
      "Data columns (total 10 columns):\n",
      "Unnamed: 0    1250 non-null int64\n",
      "Year          1250 non-null int64\n",
      "Lag1          1250 non-null float64\n",
      "Lag2          1250 non-null float64\n",
      "Lag3          1250 non-null float64\n",
      "Lag4          1250 non-null float64\n",
      "Lag5          1250 non-null float64\n",
      "Volume        1250 non-null float64\n",
      "Today         1250 non-null float64\n",
      "Direction     1250 non-null object\n",
      "dtypes: float64(7), int64(2), object(1)\n",
      "memory usage: 97.7+ KB\n"
     ]
    }
   ],
   "source": [
    "#load dataset\n",
    "smarket = pd.read_csv('/Users/YaoJunyan/Documents/ISLR_Python/Dataset/Smarket.csv')\n",
    "smarket.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <th>Year</th>\n",
       "      <th>Lag1</th>\n",
       "      <th>Lag2</th>\n",
       "      <th>Lag3</th>\n",
       "      <th>Lag4</th>\n",
       "      <th>Lag5</th>\n",
       "      <th>Volume</th>\n",
       "      <th>Today</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>1250.000000</td>\n",
       "      <td>1250.000000</td>\n",
       "      <td>1250.000000</td>\n",
       "      <td>1250.000000</td>\n",
       "      <td>1250.000000</td>\n",
       "      <td>1250.000000</td>\n",
       "      <td>1250.00000</td>\n",
       "      <td>1250.000000</td>\n",
       "      <td>1250.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>625.500000</td>\n",
       "      <td>2003.016000</td>\n",
       "      <td>0.003834</td>\n",
       "      <td>0.003919</td>\n",
       "      <td>0.001716</td>\n",
       "      <td>0.001636</td>\n",
       "      <td>0.00561</td>\n",
       "      <td>1.478305</td>\n",
       "      <td>0.003138</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>360.988227</td>\n",
       "      <td>1.409018</td>\n",
       "      <td>1.136299</td>\n",
       "      <td>1.136280</td>\n",
       "      <td>1.138703</td>\n",
       "      <td>1.138774</td>\n",
       "      <td>1.14755</td>\n",
       "      <td>0.360357</td>\n",
       "      <td>1.136334</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>2001.000000</td>\n",
       "      <td>-4.922000</td>\n",
       "      <td>-4.922000</td>\n",
       "      <td>-4.922000</td>\n",
       "      <td>-4.922000</td>\n",
       "      <td>-4.92200</td>\n",
       "      <td>0.356070</td>\n",
       "      <td>-4.922000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>313.250000</td>\n",
       "      <td>2002.000000</td>\n",
       "      <td>-0.639500</td>\n",
       "      <td>-0.639500</td>\n",
       "      <td>-0.640000</td>\n",
       "      <td>-0.640000</td>\n",
       "      <td>-0.64000</td>\n",
       "      <td>1.257400</td>\n",
       "      <td>-0.639500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>625.500000</td>\n",
       "      <td>2003.000000</td>\n",
       "      <td>0.039000</td>\n",
       "      <td>0.039000</td>\n",
       "      <td>0.038500</td>\n",
       "      <td>0.038500</td>\n",
       "      <td>0.03850</td>\n",
       "      <td>1.422950</td>\n",
       "      <td>0.038500</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>937.750000</td>\n",
       "      <td>2004.000000</td>\n",
       "      <td>0.596750</td>\n",
       "      <td>0.596750</td>\n",
       "      <td>0.596750</td>\n",
       "      <td>0.596750</td>\n",
       "      <td>0.59700</td>\n",
       "      <td>1.641675</td>\n",
       "      <td>0.596750</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>1250.000000</td>\n",
       "      <td>2005.000000</td>\n",
       "      <td>5.733000</td>\n",
       "      <td>5.733000</td>\n",
       "      <td>5.733000</td>\n",
       "      <td>5.733000</td>\n",
       "      <td>5.73300</td>\n",
       "      <td>3.152470</td>\n",
       "      <td>5.733000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        Unnamed: 0         Year         Lag1         Lag2         Lag3  \\\n",
       "count  1250.000000  1250.000000  1250.000000  1250.000000  1250.000000   \n",
       "mean    625.500000  2003.016000     0.003834     0.003919     0.001716   \n",
       "std     360.988227     1.409018     1.136299     1.136280     1.138703   \n",
       "min       1.000000  2001.000000    -4.922000    -4.922000    -4.922000   \n",
       "25%     313.250000  2002.000000    -0.639500    -0.639500    -0.640000   \n",
       "50%     625.500000  2003.000000     0.039000     0.039000     0.038500   \n",
       "75%     937.750000  2004.000000     0.596750     0.596750     0.596750   \n",
       "max    1250.000000  2005.000000     5.733000     5.733000     5.733000   \n",
       "\n",
       "              Lag4        Lag5       Volume        Today  \n",
       "count  1250.000000  1250.00000  1250.000000  1250.000000  \n",
       "mean      0.001636     0.00561     1.478305     0.003138  \n",
       "std       1.138774     1.14755     0.360357     1.136334  \n",
       "min      -4.922000    -4.92200     0.356070    -4.922000  \n",
       "25%      -0.640000    -0.64000     1.257400    -0.639500  \n",
       "50%       0.038500     0.03850     1.422950     0.038500  \n",
       "75%       0.596750     0.59700     1.641675     0.596750  \n",
       "max       5.733000     5.73300     3.152470     5.733000  "
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "smarket.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "smarket['direction_ind']=pd.factorize(smarket.Direction)[0]\n",
    "\n",
    "\n",
    "# Take a fraction of the samples where target value is 'down'\n",
    "df_no = smarket[smarket.direction_ind == 0].sample(frac=0.15)\n",
    "# Take all samples  where target value is 'up'\n",
    "df_yes = smarket[smarket.direction_ind == 1]\n",
    "df_ = df_no.append(df_yes)\n",
    "\n",
    "plt.scatter(df_[df_.Direction == 'Up'].Lag2, df_[df_.Direction == 'Up'].Volume, c='orange', marker='+',\n",
    "            linewidths=1)\n",
    "plt.scatter(df_[df_.Direction == 'Down'].Lag2, df_[df_.Direction == 'Down'].Volume, marker='o', linewidths='1',\n",
    "            edgecolors='lightblue', facecolors='white', alpha=.6)\n",
    "\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1250, 10)"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#dimension of data:\n",
    "smarket.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <th>Year</th>\n",
       "      <th>Lag1</th>\n",
       "      <th>Lag2</th>\n",
       "      <th>Lag3</th>\n",
       "      <th>Lag4</th>\n",
       "      <th>Lag5</th>\n",
       "      <th>Volume</th>\n",
       "      <th>Today</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Unnamed: 0</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.979773</td>\n",
       "      <td>0.035415</td>\n",
       "      <td>0.036022</td>\n",
       "      <td>0.038989</td>\n",
       "      <td>0.041437</td>\n",
       "      <td>0.035025</td>\n",
       "      <td>0.546348</td>\n",
       "      <td>0.035273</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Year</th>\n",
       "      <td>0.979773</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.029700</td>\n",
       "      <td>0.030596</td>\n",
       "      <td>0.033195</td>\n",
       "      <td>0.035689</td>\n",
       "      <td>0.029788</td>\n",
       "      <td>0.539006</td>\n",
       "      <td>0.030095</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Lag1</th>\n",
       "      <td>0.035415</td>\n",
       "      <td>0.029700</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-0.026294</td>\n",
       "      <td>-0.010803</td>\n",
       "      <td>-0.002986</td>\n",
       "      <td>-0.005675</td>\n",
       "      <td>0.040910</td>\n",
       "      <td>-0.026155</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Lag2</th>\n",
       "      <td>0.036022</td>\n",
       "      <td>0.030596</td>\n",
       "      <td>-0.026294</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-0.025897</td>\n",
       "      <td>-0.010854</td>\n",
       "      <td>-0.003558</td>\n",
       "      <td>-0.043383</td>\n",
       "      <td>-0.010250</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Lag3</th>\n",
       "      <td>0.038989</td>\n",
       "      <td>0.033195</td>\n",
       "      <td>-0.010803</td>\n",
       "      <td>-0.025897</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-0.024051</td>\n",
       "      <td>-0.018808</td>\n",
       "      <td>-0.041824</td>\n",
       "      <td>-0.002448</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Lag4</th>\n",
       "      <td>0.041437</td>\n",
       "      <td>0.035689</td>\n",
       "      <td>-0.002986</td>\n",
       "      <td>-0.010854</td>\n",
       "      <td>-0.024051</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-0.027084</td>\n",
       "      <td>-0.048414</td>\n",
       "      <td>-0.006900</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Lag5</th>\n",
       "      <td>0.035025</td>\n",
       "      <td>0.029788</td>\n",
       "      <td>-0.005675</td>\n",
       "      <td>-0.003558</td>\n",
       "      <td>-0.018808</td>\n",
       "      <td>-0.027084</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-0.022002</td>\n",
       "      <td>-0.034860</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Volume</th>\n",
       "      <td>0.546348</td>\n",
       "      <td>0.539006</td>\n",
       "      <td>0.040910</td>\n",
       "      <td>-0.043383</td>\n",
       "      <td>-0.041824</td>\n",
       "      <td>-0.048414</td>\n",
       "      <td>-0.022002</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.014592</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Today</th>\n",
       "      <td>0.035273</td>\n",
       "      <td>0.030095</td>\n",
       "      <td>-0.026155</td>\n",
       "      <td>-0.010250</td>\n",
       "      <td>-0.002448</td>\n",
       "      <td>-0.006900</td>\n",
       "      <td>-0.034860</td>\n",
       "      <td>0.014592</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            Unnamed: 0      Year      Lag1      Lag2      Lag3      Lag4  \\\n",
       "Unnamed: 0    1.000000  0.979773  0.035415  0.036022  0.038989  0.041437   \n",
       "Year          0.979773  1.000000  0.029700  0.030596  0.033195  0.035689   \n",
       "Lag1          0.035415  0.029700  1.000000 -0.026294 -0.010803 -0.002986   \n",
       "Lag2          0.036022  0.030596 -0.026294  1.000000 -0.025897 -0.010854   \n",
       "Lag3          0.038989  0.033195 -0.010803 -0.025897  1.000000 -0.024051   \n",
       "Lag4          0.041437  0.035689 -0.002986 -0.010854 -0.024051  1.000000   \n",
       "Lag5          0.035025  0.029788 -0.005675 -0.003558 -0.018808 -0.027084   \n",
       "Volume        0.546348  0.539006  0.040910 -0.043383 -0.041824 -0.048414   \n",
       "Today         0.035273  0.030095 -0.026155 -0.010250 -0.002448 -0.006900   \n",
       "\n",
       "                Lag5    Volume     Today  \n",
       "Unnamed: 0  0.035025  0.546348  0.035273  \n",
       "Year        0.029788  0.539006  0.030095  \n",
       "Lag1       -0.005675  0.040910 -0.026155  \n",
       "Lag2       -0.003558 -0.043383 -0.010250  \n",
       "Lag3       -0.018808 -0.041824 -0.002448  \n",
       "Lag4       -0.027084 -0.048414 -0.006900  \n",
       "Lag5        1.000000 -0.022002 -0.034860  \n",
       "Volume     -0.022002  1.000000  0.014592  \n",
       "Today      -0.034860  0.014592  1.000000  "
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#correlation matrix\n",
    "smarket.corr()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x11a791e80>]"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(smarket.Volume)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Library/Frameworks/Python.framework/Versions/3.6/lib/python3.6/site-packages/ipykernel_launcher.py:2: FutureWarning: Method .as_matrix will be removed in a future version. Use .values instead.\n",
      "  \n"
     ]
    }
   ],
   "source": [
    "log_reg = LogisticRegression()\n",
    "x=smarket[['Lag1', 'Lag2', 'Lag3', 'Lag4', 'Lag5', 'Volume']].as_matrix()\n",
    "y= smarket['Direction']\n",
    "model1= log_reg.fit(x,y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[-0.07279035 -0.04229102  0.010958    0.00921799  0.01023031  0.12793931]]\n"
     ]
    }
   ],
   "source": [
    "print(model1.coef_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimization terminated successfully.\n",
      "         Current function value: 0.691143\n",
      "         Iterations 4\n",
      "                         Results: Logit\n",
      "================================================================\n",
      "Model:              Logit            Pseudo R-squared: 0.002    \n",
      "Dependent Variable: y                AIC:              1739.8583\n",
      "Date:               2019-04-10 13:34 BIC:              1770.6436\n",
      "No. Observations:   1250             Log-Likelihood:   -863.93  \n",
      "Df Model:           5                LL-Null:          -865.59  \n",
      "Df Residuals:       1244             LLR p-value:      0.65131  \n",
      "Converged:          1.0000           Scale:            1.0000   \n",
      "No. Iterations:     4.0000                                      \n",
      "-------------------------------------------------------------------\n",
      "       Coef.     Std.Err.       z       P>|z|      [0.025    0.975]\n",
      "-------------------------------------------------------------------\n",
      "x1     0.0721      0.0501     1.4378    0.1505    -0.0262    0.1703\n",
      "x2     0.0434      0.0500     0.8678    0.3855    -0.0546    0.1415\n",
      "x3    -0.0100      0.0499    -0.1998    0.8416    -0.1077    0.0878\n",
      "x4    -0.0081      0.0499    -0.1617    0.8715    -0.1059    0.0897\n",
      "x5    -0.0097      0.0495    -0.1957    0.8449    -0.1067    0.0873\n",
      "x6    -0.0549      0.0373    -1.4730    0.1407    -0.1280    0.0181\n",
      "================================================================\n",
      "\n"
     ]
    }
   ],
   "source": [
    "#or try this way\n",
    "#statsmodels is very similar to R glm function, it can produces most of results in a single table using summary function\n",
    "logit_model =sm.Logit(pd.factorize(y)[0],x.astype(float)) #object type cannot be used here\n",
    "result= logit_model.fit()\n",
    "print(result.summary2())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                 Generalized Linear Model Regression Results                  \n",
      "==============================================================================\n",
      "Dep. Variable:                      y   No. Observations:                 1250\n",
      "Model:                            GLM   Df Residuals:                     1244\n",
      "Model Family:                Binomial   Df Model:                            5\n",
      "Link Function:                  logit   Scale:                          1.0000\n",
      "Method:                          IRLS   Log-Likelihood:                -863.93\n",
      "Date:                Wed, 10 Apr 2019   Deviance:                       1727.9\n",
      "Time:                        13:44:52   Pearson chi2:                 1.25e+03\n",
      "No. Iterations:                     4   Covariance Type:             nonrobust\n",
      "==============================================================================\n",
      "                 coef    std err          z      P>|z|      [0.025      0.975]\n",
      "------------------------------------------------------------------------------\n",
      "x1             0.0721      0.050      1.438      0.150      -0.026       0.170\n",
      "x2             0.0434      0.050      0.868      0.386      -0.055       0.141\n",
      "x3            -0.0100      0.050     -0.200      0.842      -0.108       0.088\n",
      "x4            -0.0081      0.050     -0.162      0.872      -0.106       0.090\n",
      "x5            -0.0097      0.049     -0.196      0.845      -0.107       0.087\n",
      "x6            -0.0549      0.037     -1.473      0.141      -0.128       0.018\n",
      "==============================================================================\n"
     ]
    }
   ],
   "source": [
    "model2= sm.GLM(pd.factorize(y)[0],x.astype(float), family=sm.families.Binomial())\n",
    "model2_result= model2.fit()\n",
    "print(model2_result.summary())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**train the model**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train, x_test, y_train, y_test = train_test_split(x,y, test_size =0.3, random_state= 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[-0.06756919 -0.06561412  0.04349014 -0.03971087 -0.03472513  0.09201262]]\n"
     ]
    }
   ],
   "source": [
    "model3= log_reg.fit(x_train, y_train)\n",
    "print(model3.coef_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred= model3.predict(x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[ 46 138]\n",
      " [ 50 141]]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import confusion_matrix\n",
    "confusion_matrix = confusion_matrix(y_test, y_pred)\n",
    "print(confusion_matrix)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This matrix is telling us we have 46+141 correct predictions and 50+138 incorrect predictions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "       Down       0.48      0.25      0.33       184\n",
      "         Up       0.51      0.74      0.60       191\n",
      "\n",
      "avg / total       0.49      0.50      0.47       375\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(metrics.classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The recall is the ratio **tp / (tp + fn)** where *tp* is the number of true positives and *fn* the number of false negatives. The recall is intuitively the ability of the classifier to find all the positive samples.\n",
    "The **F-beta** score can be interpreted as a weighted harmonic mean of the precision and recall, where an F-beta score reaches its best value at 1 and worst score at 0.\n",
    "The F-beta score weights the recall more than the precision by a factor of beta. beta = 1.0 means recall and precision are equally important.\n",
    "citation: https://towardsdatascience.com/building-a-logistic-regression-in-python-step-by-step-becd4d56c9c8\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy 0.49866666666666665\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import roc_curve, auc\n",
    "print(\"Accuracy\", metrics.accuracy_score(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_pred_proba = model3.predict_proba(x_test)[::,1]\n",
    "fpr, tpr, _ = metrics.roc_curve(pd.factorize(y_test)[0],  y_pred_proba)\n",
    "auc = metrics.roc_auc_score(pd.factorize(y_test)[0], y_pred_proba)\n",
    "plt.plot(fpr,tpr,label=\"data 1, auc=\"+str(auc))\n",
    "plt.plot([0, 1], [0, 1],'r--')\n",
    "plt.legend(loc=4)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "yep, both precision table and ROC curve looks pretty bad!"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## LDA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.discriminant_analysis import LinearDiscriminantAnalysis as LDA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 150,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "coefficient value [[-0.05544078 -0.0443452 ]]\n",
      "mean value [[ 0.04279022  0.03389409]\n",
      " [-0.03954635 -0.03132544]]\n"
     ]
    }
   ],
   "source": [
    "before_2005 = smarket.Year <2005\n",
    "after_2005 = smarket.Year>=2005\n",
    "X_train = smarket[before_2005][['Lag1','Lag2']]\n",
    "y_train = smarket[before_2005]['Direction']\n",
    "\n",
    "X_test = smarket[after_2005][['Lag1','Lag2']]\n",
    "y_test = smarket[after_2005]['Direction']\n",
    "lda_model =lda.fit(X_train, y_train)\n",
    "pred = lda_model.predict(X_test)\n",
    "print('coefficient value {}'.format(lda_model.coef_))\n",
    "print('mean value {}'.format(lda_model.means_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 153,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "       Down       0.50      0.32      0.39       111\n",
      "         Up       0.58      0.75      0.66       141\n",
      "\n",
      "avg / total       0.55      0.56      0.54       252\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#ROC curve for LDA model\n",
    "y_pred_proba_lda = lda_model.predict_proba(X_test)\n",
    "fpr_lda, tpr_lda, _lda = metrics.roc_curve(pd.factorize(y_test)[0],  y_pred_proba_lda[:,1])\n",
    "auc_lda = metrics.roc_auc_score(pd.factorize(y_test)[0], y_pred_proba_lda[:,1])\n",
    "plt.plot(fpr_lda,tpr_lda,label=\"data 2, auc=\"+str(auc_lda))\n",
    "plt.plot([0, 1], [0, 1],'r--')\n",
    "plt.legend(loc=4)\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "Slightly better than Logistic regression, but this is still bad"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## QDA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Prior Probabilities of groups [0.49198397 0.50801603]\n",
      "Group means [[ 0.04279022  0.03389409]\n",
      " [-0.03954635 -0.03132544]]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis\n",
    "qda = QuadraticDiscriminantAnalysis()\n",
    "qda_model = qda.fit(X_train, y_train)\n",
    "qda_pred= qda_model.predict(X_test)\n",
    "print('Prior Probabilities of groups {}'.format(qda_model.priors_))\n",
    "print('Group means {}'.format(qda_model.means_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 169,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 30,  81],\n",
       "       [ 20, 121]])"
      ]
     },
     "execution_count": 169,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "metrics.confusion_matrix(y_test, qda_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 171,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "       Down       0.60      0.27      0.37       111\n",
      "         Up       0.60      0.86      0.71       141\n",
      "\n",
      "avg / total       0.60      0.60      0.56       252\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, qda_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 174,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#ROC curve for QDA model\n",
    "y_pred_proba_qda = qda_model.predict_proba(X_test)\n",
    "fpr_qda, tpr_qda, _qda = metrics.roc_curve(pd.factorize(y_test)[0],  y_pred_proba_qda[:,1])\n",
    "auc_qda = metrics.roc_auc_score(pd.factorize(y_test)[0], y_pred_proba_qda[:,1])\n",
    "plt.plot(fpr_qda,tpr_qda,label=\"data 2, auc=\"+str(auc_qda))\n",
    "plt.plot([0, 1], [0, 1],'r--')\n",
    "plt.legend(loc=4)\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "No improvement"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## KNN with K=1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 179,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "       Down       0.43      0.39      0.41       111\n",
      "         Up       0.55      0.59      0.57       141\n",
      "\n",
      "avg / total       0.50      0.50      0.50       252\n",
      "\n",
      "[[43 68]\n",
      " [58 83]]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "classifer= KNeighborsClassifier(n_neighbors=1)\n",
    "knn_model= classifer.fit(X_train, y_train)\n",
    "knn_pred= knn_model.predict(X_test)\n",
    "print(classification_report(y_test, knn_pred))\n",
    "print(metrics.confusion_matrix(y_test, knn_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 180,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.5"
      ]
     },
     "execution_count": 180,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(43+83)/252"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## KNN with k=3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 181,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "       Down       0.47      0.43      0.45       111\n",
      "         Up       0.58      0.61      0.59       141\n",
      "\n",
      "avg / total       0.53      0.53      0.53       252\n",
      "\n",
      "[[48 63]\n",
      " [55 86]]\n"
     ]
    }
   ],
   "source": [
    "classifer2= KNeighborsClassifier(n_neighbors=3)\n",
    "knn_model2= classifer2.fit(X_train, y_train)\n",
    "knn_pred2= knn_model2.predict(X_test)\n",
    "print(classification_report(y_test, knn_pred2))\n",
    "print(metrics.confusion_matrix(y_test, knn_pred2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 182,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.5317460317460317"
      ]
     },
     "execution_count": 182,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(48+86)/252 "
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "KNN with K=3 has a better results "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## An Application to Caravan Insurance Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 212,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['MOSTYPE', 'MAANTHUI', 'MGEMOMV', 'MGEMLEEF', 'MOSHOOFD', 'MGODRK',\n",
       "       'MGODPR', 'MGODOV', 'MGODGE', 'MRELGE', 'MRELSA', 'MRELOV', 'MFALLEEN',\n",
       "       'MFGEKIND', 'MFWEKIND', 'MOPLHOOG', 'MOPLMIDD', 'MOPLLAAG', 'MBERHOOG',\n",
       "       'MBERZELF', 'MBERBOER', 'MBERMIDD', 'MBERARBG', 'MBERARBO', 'MSKA',\n",
       "       'MSKB1', 'MSKB2', 'MSKC', 'MSKD', 'MHHUUR', 'MHKOOP', 'MAUT1', 'MAUT2',\n",
       "       'MAUT0', 'MZFONDS', 'MZPART', 'MINKM30', 'MINK3045', 'MINK4575',\n",
       "       'MINK7512', 'MINK123M', 'MINKGEM', 'MKOOPKLA', 'PWAPART', 'PWABEDR',\n",
       "       'PWALAND', 'PPERSAUT', 'PBESAUT', 'PMOTSCO', 'PVRAAUT', 'PAANHANG',\n",
       "       'PTRACTOR', 'PWERKT', 'PBROM', 'PLEVEN', 'PPERSONG', 'PGEZONG',\n",
       "       'PWAOREG', 'PBRAND', 'PZEILPL', 'PPLEZIER', 'PFIETS', 'PINBOED',\n",
       "       'PBYSTAND', 'AWAPART', 'AWABEDR', 'AWALAND', 'APERSAUT', 'ABESAUT',\n",
       "       'AMOTSCO', 'AVRAAUT', 'AAANHANG', 'ATRACTOR', 'AWERKT', 'ABROM',\n",
       "       'ALEVEN', 'APERSONG', 'AGEZONG', 'AWAOREG', 'ABRAND', 'AZEILPL',\n",
       "       'APLEZIER', 'AFIETS', 'AINBOED', 'ABYSTAND', 'Purchase'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 212,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "caravan= pd.read_csv('Dataset/Caravan.csv')\n",
    "df=caravan.drop('Unnamed: 0', axis=1)\n",
    "df.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 213,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "No     5474\n",
       "Yes     348\n",
       "Name: Purchase, dtype: int64"
      ]
     },
     "execution_count": 213,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['Purchase'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 214,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 0.68090623, -0.27257995,  0.40669664, ..., -0.15062046,\n",
       "        -0.08734772, -0.11881647],\n",
       "       [ 0.99229684, -0.27257995, -0.85950007, ..., -0.15062046,\n",
       "        -0.08734772, -0.11881647],\n",
       "       [ 0.99229684, -0.27257995, -0.85950007, ..., -0.15062046,\n",
       "        -0.08734772, -0.11881647],\n",
       "       ...,\n",
       "       [ 0.68090623, -0.27257995,  0.40669664, ..., -0.15062046,\n",
       "        -0.08734772, -0.11881647],\n",
       "       [ 0.75875388, -0.27257995,  0.40669664, ..., -0.15062046,\n",
       "        -0.08734772, -0.11881647],\n",
       "       [ 0.68090623, -0.27257995,  0.40669664, ..., -0.15062046,\n",
       "        -0.08734772, -0.11881647]])"
      ]
     },
     "execution_count": 214,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#scale variables\n",
    "X= df.drop('Purchase', axis =1)\n",
    "y= df.Purchase\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "scale= StandardScaler()\n",
    "scale.fit(X)\n",
    "scale.transform(X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 215,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train =X.iloc[1000:,:]\n",
    "X_test=X.iloc[0:1000,:]\n",
    "y_train =y.iloc[1000:]\n",
    "y_test =y.iloc[0:1000]\n",
    "X_train_scale = scale.transform(X_train)\n",
    "X_test_scale= scale.transform(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 217,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[873  68]\n",
      " [ 50   9]]\n"
     ]
    }
   ],
   "source": [
    "#K=1\n",
    "knn_1= KNeighborsClassifier(n_neighbors=1)\n",
    "mod1= knn_1.fit(X_train_scale, y_train)\n",
    "knn_1_pred= mod1.predict(X_test_scale)\n",
    "print(metrics.confusion_matrix(y_test, knn_1_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 219,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.11688311688311688"
      ]
     },
     "execution_count": 219,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "9/(9+68)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 220,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[921  20]\n",
      " [ 54   5]]\n"
     ]
    }
   ],
   "source": [
    "#k=3\n",
    "knn_3= KNeighborsClassifier(n_neighbors=3)\n",
    "mod3= knn_3.fit(X_train_scale, y_train)\n",
    "knn_3_pred= mod3.predict(X_test_scale)\n",
    "print(metrics.confusion_matrix(y_test, knn_3_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 221,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.16666666666666666"
      ]
     },
     "execution_count": 221,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "5/(5+25)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 222,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[930  11]\n",
      " [ 55   4]]\n"
     ]
    }
   ],
   "source": [
    "#k=5\n",
    "knn_5= KNeighborsClassifier(n_neighbors=5)\n",
    "mod5= knn_5.fit(X_train_scale, y_train)\n",
    "knn_5_pred= mod5.predict(X_test_scale)\n",
    "print(metrics.confusion_matrix(y_test, knn_5_pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 223,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.26666666666666666"
      ]
     },
     "execution_count": 223,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "4/(4+11)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 224,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,\n",
       "          intercept_scaling=1, max_iter=100, multi_class='ovr', n_jobs=1,\n",
       "          penalty='l2', random_state=None, solver='liblinear', tol=0.0001,\n",
       "          verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 224,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Logistics Regression\n",
    "logreg =LogisticRegression()\n",
    "model4= logreg.fit(X_train, y_train)\n",
    "model4"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 228,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>No</th>\n",
       "      <th>Yes</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>No</th>\n",
       "      <td>939</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Yes</th>\n",
       "      <td>59</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      No  Yes\n",
       "No   939    2\n",
       "Yes   59    0"
      ]
     },
     "execution_count": 228,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred= model4.predict(X_test)\n",
    "pd.DataFrame(metrics.confusion_matrix(y_test, pred), index= logreg.classes_, columns=logreg.classes_ )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 229,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "         No       0.94      1.00      0.97       941\n",
      "        Yes       0.00      0.00      0.00        59\n",
      "\n",
      "avg / total       0.89      0.94      0.91      1000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 236,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>Pred</th>\n",
       "      <th>No</th>\n",
       "      <th>Yes</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>True</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>No</th>\n",
       "      <td>921</td>\n",
       "      <td>20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Yes</th>\n",
       "      <td>51</td>\n",
       "      <td>8</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Pred   No  Yes\n",
       "True          \n",
       "No    921   20\n",
       "Yes    51    8"
      ]
     },
     "execution_count": 236,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_p= model4.predict_proba(X_test)\n",
    "cm_df = pd.DataFrame({'True': y_test, 'Pred': pred_p[:,1] >0.25})\n",
    "cm_df.Pred.replace(to_replace ={True:'Yes', False: 'No'}, inplace=True)\n",
    "cm_df.groupby(['True' ,'Pred']).size().unstack('True').T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 237,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.2857142857142857"
      ]
     },
     "execution_count": 237,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "8/28"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A little improvement with k =5"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Applied exercise using Auto data set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "auto= pd.read_csv('Dataset/Auto.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 397 entries, 0 to 396\n",
      "Data columns (total 9 columns):\n",
      "mpg             397 non-null float64\n",
      "cylinders       397 non-null int64\n",
      "displacement    397 non-null float64\n",
      "horsepower      397 non-null object\n",
      "weight          397 non-null int64\n",
      "acceleration    397 non-null float64\n",
      "year            397 non-null int64\n",
      "origin          397 non-null int64\n",
      "name            397 non-null object\n",
      "dtypes: float64(3), int64(4), object(2)\n",
      "memory usage: 28.0+ KB\n"
     ]
    }
   ],
   "source": [
    "auto.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "auto.horsepower.head()\n",
    "auto['hp']= auto.horsepower.replace('?',None, regex=True).astype(float)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    397.000000\n",
       "mean      23.515869\n",
       "std        7.825804\n",
       "min        9.000000\n",
       "25%       17.500000\n",
       "50%       23.000000\n",
       "75%       29.000000\n",
       "max       46.600000\n",
       "Name: mpg, dtype: float64"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "auto.mpg.describe() "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "#so the median for mpg is 23\n",
    "mpg01= np.where(auto.mpg >23,1,0)\n",
    "mpg01 =pd.DataFrame(mpg01)\n",
    "auto ['mpg01'] = mpg01"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>mpg</th>\n",
       "      <th>cylinders</th>\n",
       "      <th>displacement</th>\n",
       "      <th>horsepower</th>\n",
       "      <th>weight</th>\n",
       "      <th>acceleration</th>\n",
       "      <th>year</th>\n",
       "      <th>origin</th>\n",
       "      <th>name</th>\n",
       "      <th>hp</th>\n",
       "      <th>mpg01</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>392</th>\n",
       "      <td>27.0</td>\n",
       "      <td>4</td>\n",
       "      <td>140.0</td>\n",
       "      <td>86</td>\n",
       "      <td>2790</td>\n",
       "      <td>15.6</td>\n",
       "      <td>82</td>\n",
       "      <td>1</td>\n",
       "      <td>ford mustang gl</td>\n",
       "      <td>86.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>393</th>\n",
       "      <td>44.0</td>\n",
       "      <td>4</td>\n",
       "      <td>97.0</td>\n",
       "      <td>52</td>\n",
       "      <td>2130</td>\n",
       "      <td>24.6</td>\n",
       "      <td>82</td>\n",
       "      <td>2</td>\n",
       "      <td>vw pickup</td>\n",
       "      <td>52.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>394</th>\n",
       "      <td>32.0</td>\n",
       "      <td>4</td>\n",
       "      <td>135.0</td>\n",
       "      <td>84</td>\n",
       "      <td>2295</td>\n",
       "      <td>11.6</td>\n",
       "      <td>82</td>\n",
       "      <td>1</td>\n",
       "      <td>dodge rampage</td>\n",
       "      <td>84.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>395</th>\n",
       "      <td>28.0</td>\n",
       "      <td>4</td>\n",
       "      <td>120.0</td>\n",
       "      <td>79</td>\n",
       "      <td>2625</td>\n",
       "      <td>18.6</td>\n",
       "      <td>82</td>\n",
       "      <td>1</td>\n",
       "      <td>ford ranger</td>\n",
       "      <td>79.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>396</th>\n",
       "      <td>31.0</td>\n",
       "      <td>4</td>\n",
       "      <td>119.0</td>\n",
       "      <td>82</td>\n",
       "      <td>2720</td>\n",
       "      <td>19.4</td>\n",
       "      <td>82</td>\n",
       "      <td>1</td>\n",
       "      <td>chevy s-10</td>\n",
       "      <td>82.0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      mpg  cylinders  displacement horsepower  weight  acceleration  year  \\\n",
       "392  27.0          4         140.0         86    2790          15.6    82   \n",
       "393  44.0          4          97.0         52    2130          24.6    82   \n",
       "394  32.0          4         135.0         84    2295          11.6    82   \n",
       "395  28.0          4         120.0         79    2625          18.6    82   \n",
       "396  31.0          4         119.0         82    2720          19.4    82   \n",
       "\n",
       "     origin             name    hp  mpg01  \n",
       "392       1  ford mustang gl  86.0      1  \n",
       "393       2        vw pickup  52.0      1  \n",
       "394       1    dodge rampage  84.0      1  \n",
       "395       1      ford ranger  79.0      1  \n",
       "396       1       chevy s-10  82.0      1  "
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "auto.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x11e59d0f0>"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter('hp', 'mpg01', data= auto)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x11e750550>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter('hp', 'cylinders', data= auto)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "y=auto['mpg01']\n",
    "x=auto.drop(['mpg','mpg01','name', 'horsepower'], axis=1)\n",
    "x_train, x_test, y_train, y_test = train_test_split(x,y, test_size =0.3, random_state= 0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## LDA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.discriminant_analysis import LinearDiscriminantAnalysis\n",
    "lda= LinearDiscriminantAnalysis()\n",
    "auto_model_lda= lda.fit(x_train, y_train)\n",
    "pred= auto_model_lda.predict(x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       0.90      0.96      0.93        57\n",
      "          1       0.97      0.90      0.93        63\n",
      "\n",
      "avg / total       0.94      0.93      0.93       120\n",
      "\n",
      "    0   1\n",
      "0  55   2\n",
      "1   6  57\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(pred, y_test))\n",
    "print(pd.DataFrame(metrics.confusion_matrix(pred, y_test),index= auto_model_lda.classes_, columns= auto_model_lda.classes_)) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## QDA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis\n",
    "qda= QuadraticDiscriminantAnalysis()\n",
    "auto_model_qda= qda.fit(x_train, y_train)\n",
    "qda_pred= auto_model_qda.predict(x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       0.90      0.96      0.93        57\n",
      "          1       0.97      0.90      0.93        63\n",
      "\n",
      "avg / total       0.94      0.93      0.93       120\n",
      "\n",
      "    0   1\n",
      "0  55   2\n",
      "1   6  57\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(qda_pred, y_test))\n",
    "print(pd.DataFrame(metrics.confusion_matrix(qda_pred, y_test),index= auto_model_lda.classes_, columns= auto_model_lda.classes_)) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Logistic Regression"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "log_reg= LogisticRegression()\n",
    "auto_model_log= log_reg.fit(x_train, y_train)\n",
    "log_pred= auto_model_log.predict(x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       0.90      0.90      0.90        61\n",
      "          1       0.90      0.90      0.90        59\n",
      "\n",
      "avg / total       0.90      0.90      0.90       120\n",
      "\n",
      "    0   1\n",
      "0  55   6\n",
      "1   6  53\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, log_pred))\n",
    "print(pd.DataFrame(metrics.confusion_matrix(log_pred, y_test), index= auto_model_log.classes_, columns = auto_model_log.classes_))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## KNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "knn_2= KNeighborsClassifier(n_neighbors=2)\n",
    "auto_model_knn2= knn_2.fit(x_train, y_train)\n",
    "knn2_pred= auto_model_knn2.predict(x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       0.81      0.92      0.86        61\n",
      "          1       0.90      0.78      0.84        59\n",
      "\n",
      "avg / total       0.86      0.85      0.85       120\n",
      "\n",
      "    0   1\n",
      "0  56  13\n",
      "1   5  46\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, knn2_pred))\n",
    "print(pd.DataFrame(metrics.confusion_matrix(knn2_pred, y_test), index= auto_model_knn2.classes_, columns = auto_model_knn2.classes_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       0.96      0.89      0.92        61\n",
      "          1       0.89      0.97      0.93        59\n",
      "\n",
      "avg / total       0.93      0.93      0.92       120\n",
      "\n",
      "    0   1\n",
      "0  54   2\n",
      "1   7  57\n"
     ]
    }
   ],
   "source": [
    "knn_3= KNeighborsClassifier(n_neighbors=3)\n",
    "auto_model_knn3= knn_3.fit(x_train, y_train)\n",
    "knn3_pred= auto_model_knn3.predict(x_test)\n",
    "print(classification_report(y_test, knn3_pred))\n",
    "print(pd.DataFrame(metrics.confusion_matrix(knn3_pred, y_test), index= auto_model_knn3.classes_, columns = auto_model_knn3.classes_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "          0       0.89      0.92      0.90        61\n",
      "          1       0.91      0.88      0.90        59\n",
      "\n",
      "avg / total       0.90      0.90      0.90       120\n",
      "\n",
      "    0   1\n",
      "0  56   7\n",
      "1   5  52\n"
     ]
    }
   ],
   "source": [
    "knn_4= KNeighborsClassifier(n_neighbors=4)\n",
    "auto_model_knn4= knn_4.fit(x_train, y_train)\n",
    "knn4_pred= auto_model_knn4.predict(x_test)\n",
    "print(classification_report(y_test, knn4_pred))\n",
    "print(pd.DataFrame(metrics.confusion_matrix(knn4_pred, y_test), index= auto_model_knn4.classes_, columns = auto_model_knn4.classes_))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Looks like K=3 works best for KNN model"
   ]
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}