ml_lesson5.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# Logistic Regression and Classification Error Metrics"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Human Activity Recognition with Smartphones"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import os\n",
    "from __future__ import print_function"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Question 1\n",
    "\n",
    "Import the data and do the following:\n",
    "\n",
    "* Examine the data types--there are many columns, so it might be wise to use value counts\n",
    "* Determine if the floating point values need to be scaled\n",
    "* Determine the breakdown of each activity\n",
    "* Encode the activity label as an integer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "\n",
    "data_path = ['../data/']\n",
    "file_name = ['Human_Activity_Recognition_Using_Smartphones_Data.csv']\n",
    "file_path = os.sep.join(data_path+file_name)\n",
    "data = pd.read_csv(file_path)\n",
    "#print(data.dtypes)\n",
    "#data.head()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "float64    561\n",
       "object       1\n",
       "dtype: int64"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.dtypes.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-1.0    561\n",
      "dtype: int64\n",
      "1.0    561\n",
      "dtype: int64\n"
     ]
    }
   ],
   "source": [
    "print(data.iloc[:,:-1].min().value_counts())\n",
    "print(data.iloc[:,:-1].max().value_counts())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LAYING                1944\n",
       "STANDING              1906\n",
       "SITTING               1777\n",
       "WALKING               1722\n",
       "WALKING_UPSTAIRS      1544\n",
       "WALKING_DOWNSTAIRS    1406\n",
       "Name: Activity, dtype: int64"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.Activity.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "969     5\n",
       "6674    1\n",
       "873     2\n",
       "8598    1\n",
       "8297    1\n",
       "2100    1\n",
       "7521    2\n",
       "4231    1\n",
       "6936    4\n",
       "9462    0\n",
       "Name: Activity, dtype: int64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.preprocessing import LabelEncoder\n",
    "import numpy as np\n",
    "\n",
    "le =  LabelEncoder()\n",
    "data['Activity'] = le.fit_transform(data['Activity'])\n",
    "data['Activity'].sample(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Question 2\n",
    "\n",
    "* Calculate the correlations between the dependent variables.\n",
    "* Create a histogram of the correlation values\n",
    "* Identify those that are most correlated (either positively or negatively)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "feature_columns = data.columns[:-1]\n",
    "corr_values = data[feature_columns].corr()\n",
    "#corr_values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>feature_1</th>\n",
       "      <th>feature_2</th>\n",
       "      <th>correlation</th>\n",
       "      <th>abs_correlation</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>tBodyAcc-mean()-X</td>\n",
       "      <td>tBodyAcc-mean()-Y</td>\n",
       "      <td>0.128037</td>\n",
       "      <td>0.128037</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>tBodyAcc-mean()-X</td>\n",
       "      <td>tBodyAcc-mean()-Z</td>\n",
       "      <td>-0.230302</td>\n",
       "      <td>0.230302</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>tBodyAcc-mean()-X</td>\n",
       "      <td>tBodyAcc-std()-X</td>\n",
       "      <td>0.004590</td>\n",
       "      <td>0.004590</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>tBodyAcc-mean()-X</td>\n",
       "      <td>tBodyAcc-std()-Y</td>\n",
       "      <td>-0.016785</td>\n",
       "      <td>0.016785</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>tBodyAcc-mean()-X</td>\n",
       "      <td>tBodyAcc-std()-Z</td>\n",
       "      <td>-0.036071</td>\n",
       "      <td>0.036071</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           feature_1          feature_2  correlation  abs_correlation\n",
       "0  tBodyAcc-mean()-X  tBodyAcc-mean()-Y     0.128037         0.128037\n",
       "1  tBodyAcc-mean()-X  tBodyAcc-mean()-Z    -0.230302         0.230302\n",
       "2  tBodyAcc-mean()-X   tBodyAcc-std()-X     0.004590         0.004590\n",
       "3  tBodyAcc-mean()-X   tBodyAcc-std()-Y    -0.016785         0.016785\n",
       "4  tBodyAcc-mean()-X   tBodyAcc-std()-Z    -0.036071         0.036071"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "trail_index = np.tril_indices_from(corr_values)\n",
    "\n",
    "trail_index\n",
    "for coord in zip(*trail_index):\n",
    "    corr_values.iloc[coord[0],coord[1]] =np.NaN\n",
    "#corr_values = (corr_values.stack().to_frame().reset_index())   \n",
    "corr_values = (corr_values.stack().to_frame().reset_index().rename(columns = {'level_0':'feature_1',\n",
    "                                            'level_1':'feature_2',0:'correlation'}))\n",
    "corr_values['abs_correlation'] = corr_values.correlation.abs()\n",
    "corr_values.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[Text(0,0.5,'Frequency'),\n",
       " Text(0.5,0,'Absolute Correlation'),\n",
       " Text(0.5,1,'Correlation')]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAnsAAAJmCAYAAAAkfEbSAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f9d9e0c0710>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "%matplotlib inline\n",
    "fig = plt.figure(figsize=(10,10))\n",
    "sns.set_style('white')\n",
    "sns.set_context('talk')\n",
    "sns.set_palette('dark')\n",
    "ax = corr_values.abs_correlation.hist(bins=50)\n",
    "ax.set(xlabel='Absolute Correlation',ylabel='Frequency',title='Correlation')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>feature_1</th>\n",
       "      <th>feature_2</th>\n",
       "      <th>correlation</th>\n",
       "      <th>abs_correlation</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>36294</th>\n",
       "      <td>tGravityAcc-arCoeff()-Y,1</td>\n",
       "      <td>tGravityAcc-arCoeff()-Y,2</td>\n",
       "      <td>-0.991609</td>\n",
       "      <td>0.991609</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27793</th>\n",
       "      <td>tGravityAcc-max()-Z</td>\n",
       "      <td>angle(Z,gravityMean)</td>\n",
       "      <td>-0.992689</td>\n",
       "      <td>0.992689</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22657</th>\n",
       "      <td>tGravityAcc-mean()-Y</td>\n",
       "      <td>angle(Y,gravityMean)</td>\n",
       "      <td>-0.993425</td>\n",
       "      <td>0.993425</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39225</th>\n",
       "      <td>tGravityAcc-arCoeff()-Z,3</td>\n",
       "      <td>tGravityAcc-arCoeff()-Z,4</td>\n",
       "      <td>-0.994267</td>\n",
       "      <td>0.994267</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38739</th>\n",
       "      <td>tGravityAcc-arCoeff()-Z,2</td>\n",
       "      <td>tGravityAcc-arCoeff()-Z,3</td>\n",
       "      <td>-0.994628</td>\n",
       "      <td>0.994628</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23176</th>\n",
       "      <td>tGravityAcc-mean()-Z</td>\n",
       "      <td>angle(Z,gravityMean)</td>\n",
       "      <td>-0.994764</td>\n",
       "      <td>0.994764</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38252</th>\n",
       "      <td>tGravityAcc-arCoeff()-Z,1</td>\n",
       "      <td>tGravityAcc-arCoeff()-Z,2</td>\n",
       "      <td>-0.995195</td>\n",
       "      <td>0.995195</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                       feature_1                  feature_2  correlation  \\\n",
       "36294  tGravityAcc-arCoeff()-Y,1  tGravityAcc-arCoeff()-Y,2    -0.991609   \n",
       "27793        tGravityAcc-max()-Z       angle(Z,gravityMean)    -0.992689   \n",
       "22657       tGravityAcc-mean()-Y       angle(Y,gravityMean)    -0.993425   \n",
       "39225  tGravityAcc-arCoeff()-Z,3  tGravityAcc-arCoeff()-Z,4    -0.994267   \n",
       "38739  tGravityAcc-arCoeff()-Z,2  tGravityAcc-arCoeff()-Z,3    -0.994628   \n",
       "23176       tGravityAcc-mean()-Z       angle(Z,gravityMean)    -0.994764   \n",
       "38252  tGravityAcc-arCoeff()-Z,1  tGravityAcc-arCoeff()-Z,2    -0.995195   \n",
       "\n",
       "       abs_correlation  \n",
       "36294         0.991609  \n",
       "27793         0.992689  \n",
       "22657         0.993425  \n",
       "39225         0.994267  \n",
       "38739         0.994628  \n",
       "23176         0.994764  \n",
       "38252         0.995195  "
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "corr_values = corr_values.sort_values('correlation',ascending=False).query('abs_correlation > .8')\n",
    "corr_values.tail(7)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Question 3\n",
    "\n",
    "* Split the data into train and test data sets. This can be done using any method, but consider using Scikit-learn's `StratifiedShuffleSplit` to maintain the same ratio of predictor classes.\n",
    "* Regardless of methods used to split the data, compare the ratio of classes in both the train and test splits.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.model_selection import StratifiedShuffleSplit\n",
    "sss = StratifiedShuffleSplit(n_splits=1 ,test_size = .3,random_state = 42)\n",
    "\n",
    "train_index, test_index = next(sss.split(data[feature_columns],data.Activity))\n",
    "X_train = data.loc[train_index,feature_columns]\n",
    "Y_train = data.loc[train_index,'Activity']\n",
    "X_test = data.loc[test_index,feature_columns]\n",
    "Y_test = data.loc[test_index,'Activity']\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    0.188792\n",
       "2    0.185046\n",
       "1    0.172562\n",
       "3    0.167152\n",
       "5    0.149951\n",
       "4    0.136496\n",
       "Name: Activity, dtype: float64"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Y_train.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    0.188673\n",
       "2    0.185113\n",
       "1    0.172492\n",
       "3    0.167314\n",
       "5    0.149838\n",
       "4    0.136570\n",
       "Name: Activity, dtype: float64"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Y_test.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Question 4\n",
    "\n",
    "* Fit a logistic regression model without any regularization using all of the features. Be sure to read the documentation about fitting a multi-class model so you understand the coefficient output. Store the model.\n",
    "* Using cross validation to determine the hyperparameters, fit models using L1, and L2 regularization. Store each of these models as well. Note the limitations on multi-class models, solvers, and regularizations. The regularized models, in particular the L1 model, will probably take a while to fit."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "\n",
    "lr = LogisticRegression().fit(X_train, Y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "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": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LogisticRegressionCV(Cs=10, class_weight=None, cv=4, dual=False,\n",
       "           fit_intercept=True, intercept_scaling=1.0, max_iter=100,\n",
       "           multi_class='ovr', n_jobs=1, penalty='l1', random_state=None,\n",
       "           refit=True, scoring=None, solver='liblinear', tol=0.0001,\n",
       "           verbose=0)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.linear_model import LogisticRegressionCV\n",
    "lr_l1 = LogisticRegressionCV(Cs=10,cv=4,penalty='l1',solver='liblinear')\n",
    "lr_l1.fit(X_train, Y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LogisticRegressionCV(Cs=10, class_weight=None, cv=4, dual=False,\n",
       "           fit_intercept=True, intercept_scaling=1.0, max_iter=100,\n",
       "           multi_class='ovr', n_jobs=1, penalty='l2', random_state=None,\n",
       "           refit=True, scoring=None, solver='lbfgs', tol=0.0001, verbose=0)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lr_l2 = LogisticRegressionCV(Cs=10,cv=4,penalty='l2')\n",
    "lr_l2.fit(X_train, Y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Question 5\n",
    "\n",
    "Compare the magnitudes of the coefficients for each of the models. If one-vs-rest fitting was used, each set of coefficients can be plotted separately. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[-0.08750646  0.01449275  0.03018717 ...,  1.59910474 -0.3743129\n",
      "  -0.18830925]\n",
      " [-0.44522584 -0.20104648  0.06319637 ..., -2.47309446 -0.74039698\n",
      "  -0.2980224 ]\n",
      " [ 0.25796347  0.06089227  0.24590906 ..., -0.85461228  1.77145495\n",
      "   0.41692013]\n",
      " [-0.16928095 -0.02350969  0.02234019 ..., -0.31314455 -0.24290354\n",
      "   0.0650486 ]\n",
      " [ 0.76561082  0.1079669   0.0589804  ..., -0.71269424 -0.27513866\n",
      "  -0.04236269]\n",
      " [-0.08008384 -0.24719318 -0.40252633 ...,  0.4163113   0.87672796\n",
      "   0.43881539]]\n",
      "[[-0.32154206  0.          0.         ...,  4.40173849  0.          0.        ]\n",
      " [ 0.          0.          0.         ..., -8.00980507  0.          0.        ]\n",
      " [ 0.          0.          0.         ...,  0.          0.          0.        ]\n",
      " [-0.88766065  0.          0.         ..., -0.33829697 -0.41244606  0.        ]\n",
      " [ 4.43169957  0.          0.         ..., -1.06616396  0.          0.        ]\n",
      " [ 0.          0.          0.         ...,  0.35470718  1.94125999\n",
      "   1.159715  ]]\n",
      "[[-0.2357898   0.05838939  0.04292471 ...,  3.89739399 -0.73682084\n",
      "  -0.44975055]\n",
      " [-0.71826558 -0.38848507 -0.15463045 ..., -3.83681992 -0.39478075\n",
      "  -0.24735289]\n",
      " [ 0.54756511  0.13137599  0.38900577 ..., -1.00742671  2.2479122\n",
      "   0.57965073]\n",
      " [-0.09065598 -0.13244637  0.0201761  ..., -0.69541535 -0.63829691\n",
      "   0.12475737]\n",
      " [ 2.74625204  0.26740292 -0.10008599 ..., -2.20435044 -1.13676592\n",
      "  -0.2292703 ]\n",
      " [ 0.07409407 -0.55855234 -1.0689597  ...,  0.92475466  1.97684503\n",
      "   1.02682255]]\n"
     ]
    }
   ],
   "source": [
    "models = [lr,lr_l1,lr_l2]\n",
    "labels = ['lr','lr_l1','lr_l2']\n",
    "\n",
    "coefficients = list()\n",
    "lr_l1.coef_\n",
    "for lab,mod in zip(labels,models):\n",
    "    coeff = mod.coef_\n",
    "    coeff_labels = pd.MultiIndex(levels = [[lab],[0,1,2,3,4,5]],\n",
    "                                labels = [[0,0,0,0,0,0],[0,1,2,3,4,5]],names=['first','second'])\n",
    "    coefficients.append(pd.DataFrame(coeff.T,columns=coeff_labels))\n",
    "    print(coeff)\n",
    "    \n",
    "coefficients = pd.concat(coefficients,axis=1)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "## Question 6\n",
    "\n",
    "* Predict and store the class for each model.\n",
    "* Also store the probability for the predicted class for each model. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>lr</th>\n",
       "      <th>lr_l1</th>\n",
       "      <th>lr_l2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   lr  lr_l1  lr_l2\n",
       "0   3      3      3\n",
       "1   5      5      5\n",
       "2   3      3      3\n",
       "3   1      1      1\n",
       "4   0      0      0\n",
       "5   5      5      5"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "levels = ['lr','lr_l1','lr_l2']\n",
    "models = [lr , lr_l1, lr_l2]\n",
    "\n",
    "y_pred = list()\n",
    "y_prob = list()\n",
    "\n",
    "for lab,mod in zip(levels,models):\n",
    "    y_pred.append( pd.Series(mod.predict(X_test), name = lab))\n",
    "    y_prob.append( pd.Series(mod.predict_proba(X_test).max(axis=1), name = lab))\n",
    "\n",
    "y_pred =pd.concat(y_pred,axis=1)\n",
    "y_prob =pd.concat(y_prob,axis=1)\n",
    "y_pred.head(6)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Question 7\n",
    "\n",
    "For each model, calculate the following error metrics: \n",
    "\n",
    "* accuracy\n",
    "* precision\n",
    "* recall\n",
    "* fscore\n",
    "* confusion matrix\n",
    "\n",
    "Decide how to combine the multi-class metrics into a single value for each model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>lr</th>\n",
       "      <th>lr_l1</th>\n",
       "      <th>lr_l2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>accuracy</th>\n",
       "      <td>0.984142</td>\n",
       "      <td>0.983819</td>\n",
       "      <td>0.984466</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>auc</th>\n",
       "      <td>0.990384</td>\n",
       "      <td>0.990144</td>\n",
       "      <td>0.990548</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>fscore</th>\n",
       "      <td>0.984143</td>\n",
       "      <td>0.983815</td>\n",
       "      <td>0.984466</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>precision</th>\n",
       "      <td>0.984144</td>\n",
       "      <td>0.983835</td>\n",
       "      <td>0.984471</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>recall</th>\n",
       "      <td>0.984142</td>\n",
       "      <td>0.983819</td>\n",
       "      <td>0.984466</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                 lr     lr_l1     lr_l2\n",
       "accuracy   0.984142  0.983819  0.984466\n",
       "auc        0.990384  0.990144  0.990548\n",
       "fscore     0.984143  0.983815  0.984466\n",
       "precision  0.984144  0.983835  0.984471\n",
       "recall     0.984142  0.983819  0.984466"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.metrics import precision_recall_fscore_support as score\n",
    "from sklearn.metrics import accuracy_score,roc_auc_score,confusion_matrix\n",
    "\n",
    "from sklearn.preprocessing import label_binarize\n",
    "#label_binarize(Y_test, classes=[0,1,2,3,4,5])\n",
    "cm = dict()\n",
    "metrics = list()\n",
    "\n",
    "for lab in levels:\n",
    "    precision, recall, fscore, _ = score(Y_test,y_pred[lab],average='weighted')\n",
    "    accuracy = accuracy_score(Y_test,y_pred[lab])\n",
    "    auc = roc_auc_score(label_binarize(Y_test,[0,1,2,3,4,5]), label_binarize(y_pred[lab],[0,1,2,3,4,5]), average = 'weighted')\n",
    "    \n",
    "    cm[lab] = confusion_matrix(Y_test, y_pred[lab])\n",
    "    metrics.append(pd.Series({'precision':precision,\n",
    "                   'recall':recall,\n",
    "                   'fscore':fscore,\n",
    "                   'accuracy':accuracy,\n",
    "                   'auc':auc},name=lab))\n",
    "    \n",
    "metrics = pd.concat(metrics,axis=1)\n",
    "metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Question 8\n",
    "\n",
    "Display or plot the confusion matrix for each model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[583,   0,   0,   0,   0,   0],\n",
       "       [  0, 512,  21,   0,   0,   0],\n",
       "       [  0,  22, 550,   0,   0,   0],\n",
       "       [  0,   0,   0, 515,   1,   1],\n",
       "       [  0,   0,   0,   1, 420,   1],\n",
       "       [  0,   0,   0,   1,   1, 461]])"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cm['lr']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA0wAAAK8CAYAAADceD8aAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f9d880c9978>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, axList = plt.subplots(nrows=2,ncols=2)\n",
    "axList = axList.flatten()\n",
    "\n",
    "fig.set_size_inches(12,10)\n",
    "axList[-1].axis('off')\n",
    "\n",
    "for axes, lab in zip(axList[:-1],levels):\n",
    "    sns.heatmap(cm[lab],ax = axes, annot=True,fmt='d')\n",
    "    axes.set(title=lab)\n",
    "    \n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Question 9\n",
    " Identify highly correlated columns and drop those columns before building models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.feature_selection import SelectKBest\n",
    "from sklearn.feature_selection import chi2\n",
    "from sklearn.feature_selection import VarianceThreshold\n",
    "\n",
    "#threshold with .7\n",
    "\n",
    "sel = VarianceThreshold(threshold=(.7 * (1 - .7)))\n",
    "\n",
    "data2 = pd.concat([X_train,X_test])\n",
    "data_new = pd.DataFrame(sel.fit_transform(data2))\n",
    "\n",
    "\n",
    "data_y = pd.concat([Y_train,Y_test])\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "X_new,X_test_new = train_test_split(data_new)\n",
    "Y_new,Y_test_new = train_test_split(data_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "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": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "LR = LogisticRegression()\n",
    "LR.fit(X_new,Y_new)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr_l1_new = LogisticRegressionCV(Cs=10,cv=4,penalty='l1',solver='liblinear')\n",
    "lr_l1_new.fit(X_new,Y_new)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "lr_l2_new = LogisticRegressionCV(Cs=10,cv=4,penalty='l2')\n",
    "lr_l2_new.fit(X_new,Y_new)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Try with different solvers like ‘newton-cg’, ‘lbfgs’, ‘sag’, ‘saga’ and give your observations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Question 10\n",
    "\n",
    "Compare the magnitudes of the coefficients for each of the models. If one-vs-rest fitting was used, each set of coefficients can be plotted separately. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Prepare six separate plots for each of the multi-class coefficients."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Question 11\n",
    "\n",
    "* Predict and store the class for each model.\n",
    "* Also store the probability for the predicted class for each model. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Question 12\n",
    "\n",
    "For each model, calculate the following error metrics: \n",
    "\n",
    "* accuracy\n",
    "* precision\n",
    "* recall\n",
    "* fscore\n",
    "* confusion matrix\n",
    "\n",
    "Decide how to combine the multi-class metrics into a single value for each model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Question 13\n",
    "\n",
    "Display or plot the confusion matrix for each model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}