linear_regressions.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 176,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import pickle\n",
    "import csv\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "df = pd.read_csv('data/records_4000.tsv', sep='\\t', index_col = 1)\n",
    "df['genre'] = df['genre'].str[:-1]\n",
    "df=df.fillna(0)\n",
    "df=df.replace([np.inf, -np.inf], np.nan).dropna(axis=1)\n",
    "# Format of genre is \"Rap\\n\" with the return character\n",
    "def create_train_validation_test(df, genre):\n",
    "    # 4000 songs from genre\n",
    "    labels = list(set([label.strip() for label in y]))\n",
    "    otherlabels=list(set(labels)-set([genre]))\n",
    "     #df['genre'] = df['genre'].str[:].replace([genre], '1')\n",
    "    #df['genre'] = df['genre'].str[:].replace(otherlabels, '0')\n",
    "    df_poprock = df.loc[df['genre'] == genre]\n",
    "\n",
    "    # 4000 songs from not genre\n",
    "    df_non_poprock = df.loc[df['genre'] != genre].sample(n=4000)\n",
    "\n",
    "    # Divide pop rock into train and test\n",
    "    train_poprock, test_poprock = train_test_split(df_poprock, test_size=0.2)\n",
    "\n",
    "    # Divide non pop rock into train and test\n",
    "    train_non_poprock, test_non_poprock = train_test_split(df_non_poprock, test_size = 0.2)\n",
    "\n",
    "    # Combine pop-rock and non-poprock into train/test\n",
    "    train = train_poprock.append(train_non_poprock)\n",
    "    test = test_poprock.append(test_non_poprock)\n",
    "\n",
    "    train, validation = train_test_split(train, test_size=0.2)\n",
    "\n",
    "    # training, validation, and test data\n",
    "\n",
    "    X_train = train.drop(['genre'], axis = 1)\n",
    "    y_train = train[['genre']].replace([genre], '1').replace(otherlabels, '-1')\n",
    "\n",
    "    X_validation = validation.drop(['genre'], axis = 1)\n",
    "    y_validation = validation[['genre']].replace([genre], '1').replace(otherlabels, '-1')\n",
    "\n",
    "    X_test = test.drop(['genre'], axis = 1)\n",
    "    y_test = test[['genre']].replace([genre], '1').replace(otherlabels, '-1')\n",
    "\n",
    "    return X_train, y_train, X_validation, y_validation, X_test, y_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 177,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "y=df['genre']\n",
    "labels = list(set([label.strip() for label in y]))\n",
    "X_train, y_train, X_validation, y_validation, X_test, y_test=create_train_validation_test(df,'Rap')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 178,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn import datasets, linear_model\n",
    "from sklearn.metrics import mean_squared_error, r2_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 324,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "regr = linear_model.LinearRegression()\n",
    "\n",
    "x = pd.concat([X_train,X_test]) #returns a numpy array\n",
    "min_max_scaler = preprocessing.MinMaxScaler()\n",
    "min_max_scaler.fit(x)\n",
    "X_train =min_max_scaler.transform(X_train)\n",
    "X_test =min_max_scaler.transform(X_test)\n",
    "regr.fit(X_train, y_train)\n",
    "# Make predictions using the testing set\n",
    "pred = regr.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 325,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def match(label,pred):\n",
    "    match=0\n",
    "    for i in range(0,len(label)):\n",
    "        if(label[i]*pred[i]>0):\n",
    "            match+=1;\n",
    "    return match*1.0/len(label)\n",
    "    #print \"accuracy: \"+str(match*1.0/len(label))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 326,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.679375"
      ]
     },
     "execution_count": 326,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "y_test_list=[int(y[0]) for y in list(y_test.as_matrix())]\n",
    "pred_list=[p[0] for p in pred]\n",
    "match(y_test_list,pred_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 327,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "accuracy for Reggae: 0.801875\n",
      "accuracy for Pop_Rock: 0.740625\n",
      "accuracy for Country: 0.731875\n",
      "accuracy for Jazz: 0.69625\n",
      "accuracy for Vocal: 0.7575\n",
      "accuracy for New Age: 0.77375\n",
      "accuracy for Latin: 0.65625\n",
      "accuracy for Rap: 0.784375\n",
      "accuracy for RnB: 0.63875\n",
      "accuracy for International: 0.5725\n",
      "accuracy for Blues: 0.63125\n",
      "accuracy for Electronic: 0.703125\n",
      "accuracy for Folk: 0.66625\n",
      "Avergae Accuracy: 0.704182692308\n"
     ]
    }
   ],
   "source": [
    "accus=[]\n",
    "coefs=np.zeros((13,28))\n",
    "count=0\n",
    "for labelname in labels:\n",
    "    y=df['genre']\n",
    "    labels = list(set([label.strip() for label in y]))\n",
    "    X_train, y_train, X_validation, y_validation, X_test, y_test=create_train_validation_test(df,labelname)\n",
    "    regr = linear_model.LinearRegression()\n",
    "    regr.fit(X_train, y_train)\n",
    "    pred = regr.predict(X_test)\n",
    "    y_test_list=[int(y[0]) for y in list(y_test.as_matrix())]\n",
    "    pred_list=[p[0] for p in pred]\n",
    "    accu=match(y_test_list,pred_list)\n",
    "    #print(\"Mean squared error: %.2f\"% mean_squared_error(y_test_list,pred_list))\n",
    "    print \"accuracy for \"+labelname+\": \"+str(accu)\n",
    "    accus+=[accu]\n",
    "    coefs[count]=regr.coef_\n",
    "    count+=1\n",
    "print \"Avergae Accuracy: \"+str(sum(accus)*1.0/len(accus))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 310,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "accuracy for Reggae: 0.781875\n",
      "accuracy for Pop_Rock: 0.701875\n",
      "accuracy for Country: 0.701875\n",
      "accuracy for Jazz: 0.673125\n",
      "accuracy for Vocal: 0.7125\n",
      "accuracy for New Age: 0.773125\n",
      "accuracy for Latin: 0.6425\n",
      "accuracy for Rap: 0.766875\n",
      "accuracy for RnB: 0.645\n",
      "accuracy for International: 0.566875\n",
      "accuracy for Blues: 0.593125\n",
      "accuracy for Electronic: 0.68375\n",
      "accuracy for Folk: 0.65875\n",
      "Avergae Accuracy: 0.684711538462\n"
     ]
    }
   ],
   "source": [
    "accus=[]\n",
    "coefs=np.zeros((13,28))\n",
    "count=0\n",
    "for labelname in labels:\n",
    "    y=df['genre']\n",
    "    labels = list(set([label.strip() for label in y]))\n",
    "    X_train, y_train, X_validation, y_validation, X_test, y_test=create_train_validation_test(df,labelname)\n",
    "    regr = linear_model.HuberRegressor()\n",
    "    regr.fit(X_train, y_train.values.flatten())\n",
    "    pred = regr.predict(X_test)\n",
    "    y_test_list=[int(y[0]) for y in list(y_test.as_matrix())]\n",
    "    pred_list=list(pred)\n",
    "    accu=match(y_test_list,pred_list)\n",
    "    print \"accuracy for \"+labelname+\": \"+str(accu)\n",
    "    accus+=[accu]\n",
    "    coefs[count]=regr.coef_\n",
    "    count+=1\n",
    "print \"Avergae Accuracy: \"+str(sum(accus)*1.0/len(accus))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 204,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.decomposition import PCA\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 221,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "accuracy for Reggae: 0.735\n",
      "accuracy for Pop_Rock: 0.641875\n",
      "accuracy for Country: 0.660625\n",
      "accuracy for Jazz: 0.64875\n",
      "accuracy for Vocal: 0.70125\n",
      "accuracy for New Age: 0.725\n",
      "accuracy for Latin: 0.6425\n",
      "accuracy for Rap: 0.7275\n",
      "accuracy for RnB: 0.594375\n",
      "accuracy for International: 0.554375\n",
      "accuracy for Blues: 0.565\n",
      "accuracy for Electronic: 0.6575\n",
      "accuracy for Folk: 0.64125\n",
      "Avergae Accuracy: 0.653461538462\n"
     ]
    }
   ],
   "source": [
    "accus=[]\n",
    "for labelname in labels:\n",
    "    y=df['genre']\n",
    "    labels = list(set([label.strip() for label in y]))\n",
    "    X_train, y_train, X_validation, y_validation, X_test, y_test=create_train_validation_test(df,labelname)\n",
    "    pca = PCA(n_components=10)\n",
    "    pca.fit(pd.concat([X_train,X_test]))\n",
    "    newX_train = pca.transform(X_train)\n",
    "    newX_test = pca.transform(X_test)\n",
    "    \n",
    "    regr = linear_model.LinearRegression()\n",
    "    regr.fit(newX_train, y_train)\n",
    "    pred = regr.predict(newX_test)\n",
    "    y_test_list=[int(y[0]) for y in list(y_test.as_matrix())]\n",
    "    pred_list=[p[0] for p in pred]\n",
    "    accu=match(y_test_list,pred_list)\n",
    "    accus+=[accu]\n",
    "    print \"accuracy for \"+labelname+\": \"+str(accu)\n",
    "print \"Avergae Accuracy: \"+str(sum(accus)*1.0/len(accus))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 230,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[  3.51310852e-04,   1.17641409e-02,   2.27108703e-01,\n",
       "          7.99521519e-03,   9.14736797e-02,   3.41933050e-01,\n",
       "          1.01169659e+00,   5.26338669e-03,  -3.85974865e-03,\n",
       "         -5.12532389e-02,  -9.35131661e-03,   8.18391320e-03,\n",
       "         -1.38996036e-01,  -2.23554996e-01,   2.72010018e-03,\n",
       "         -1.11721978e-02,  -2.14284796e-01,  -3.98016241e-02,\n",
       "          7.54676627e+00,  -6.82089034e+00,   6.26358547e-02,\n",
       "         -3.18260123e-02,  -2.32523928e-02,   2.30875207e-02,\n",
       "         -8.57922811e-01,   1.76932223e+00,   2.14968290e-01,\n",
       "          1.01436481e-01]])"
      ]
     },
     "execution_count": 230,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "numpy.zeros"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 231,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1, 28)"
      ]
     },
     "execution_count": 231,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "regr.coef_.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 238,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "coefs=np.zeros((13,28))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 241,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[  2.46680586e-03,   4.66439122e-03,   5.02542265e-02,\n",
       "         -2.09334246e-02,  -1.31812037e-01,  -9.18913991e-02,\n",
       "         -1.47577458e-01,   9.52352299e-03,  -1.31821106e-03,\n",
       "          2.49212499e-02,  -1.03167228e-02,  -8.42444115e-04,\n",
       "          1.28407615e-01,  -1.95177509e-01,   3.38145114e-03,\n",
       "          6.66364086e-03,   2.54889633e-01,   1.31451551e-01,\n",
       "          9.87301477e+00,  -3.67783341e+00,  -1.05140994e-01,\n",
       "         -8.83271359e-02,   1.42104155e-02,   2.60074642e-02,\n",
       "          1.01111989e+00,  -3.59623854e-01,  -1.06644851e-01,\n",
       "          1.88151849e-01],\n",
       "       [  1.33665208e-03,  -2.48116158e-03,  -2.12279539e-01,\n",
       "          2.28659858e-03,  -1.60600581e-02,   2.19927119e-02,\n",
       "          6.29045875e-02,  -3.38207960e-03,   4.70676857e-03,\n",
       "         -2.62707122e-02,  -2.98032474e-03,   2.89563019e-03,\n",
       "         -1.45608225e-01,   2.58983906e-01,  -2.06154372e-03,\n",
       "         -4.32566120e-03,  -1.18212082e-01,  -3.66425547e-02,\n",
       "         -5.37743441e+00,   2.69632775e+00,   1.71700516e-01,\n",
       "          1.30549591e-01,  -6.89957934e-03,  -8.51179117e-03,\n",
       "          4.35009606e-01,   7.30617577e-01,   1.33908411e-01,\n",
       "         -2.34313999e-01],\n",
       "       [ -1.40258411e-04,   1.07531815e-02,   1.77453514e-01,\n",
       "          1.11138291e-02,  -1.66064017e-02,   4.09595628e-01,\n",
       "          8.56742384e-01,  -4.48616350e-03,  -8.83698734e-03,\n",
       "          7.15868396e-03,  -3.37629205e-03,   3.01838838e-03,\n",
       "          4.64044033e-02,  -7.07629216e-02,   3.77830007e-03,\n",
       "         -5.90945470e-03,  -2.00775082e-02,   8.52691212e-03,\n",
       "         -7.45658915e-01,  -2.38257742e-02,   5.30929856e-02,\n",
       "         -2.58888085e-02,  -4.30002613e-03,  -4.69322857e-03,\n",
       "         -5.66657760e-02,  -7.44425783e-01,  -4.85479881e-02,\n",
       "          7.37436453e-02],\n",
       "       [ -1.07985171e-03,  -4.50481737e-03,  -6.86933880e-02,\n",
       "         -5.72737508e-04,   1.68601025e-02,  -8.61041253e-02,\n",
       "         -4.15010808e-01,  -2.57457395e-02,  -1.71985004e-03,\n",
       "          6.05872512e-03,  -9.44116161e-03,  -2.87826202e-03,\n",
       "          1.20290488e-01,  -1.84233598e-01,  -1.35116679e-03,\n",
       "         -4.22362916e-04,  -2.47492517e-02,   4.56037874e-02,\n",
       "         -9.83472150e+00,   6.64595022e-01,  -2.83089172e-02,\n",
       "         -3.30822791e-02,   8.36957726e-03,  -9.08638163e-03,\n",
       "         -1.04733928e+00,   3.34315481e-01,  -8.15063968e-02,\n",
       "          1.45697035e-01],\n",
       "       [ -2.60418253e-03,  -2.91326856e-03,   2.45341827e-01,\n",
       "         -5.48492385e-03,   1.14763299e-01,   1.08049942e-02,\n",
       "         -6.56811650e-01,  -1.93364316e-02,   4.60451213e-03,\n",
       "          2.67551431e-02,   3.97956039e-03,   8.04567562e-03,\n",
       "          5.36066574e-02,  -1.01468089e-01,   1.36995065e-03,\n",
       "          3.50917330e-03,  -1.67230117e-01,  -1.60421970e-01,\n",
       "          1.15868931e+00,  -1.97278331e+00,   1.01373146e-02,\n",
       "          1.24265264e-01,  -4.04411187e-03,  -6.46732181e-03,\n",
       "         -1.38240748e+00,   2.15365166e-01,  -3.68798413e-02,\n",
       "          6.41598019e-02],\n",
       "       [  1.08227984e-03,  -3.15830773e-03,  -2.88078415e-01,\n",
       "          1.54109514e-02,  -5.81324512e-02,  -1.84450831e-01,\n",
       "          6.74002913e-01,   1.93880420e-02,  -8.39833195e-03,\n",
       "          8.22371718e-04,   1.09885669e-03,  -1.33808957e-02,\n",
       "         -8.34436309e-02,   1.65278583e-01,  -3.05838915e-03,\n",
       "         -6.28529259e-03,  -1.00080794e-01,   9.73469912e-02,\n",
       "         -1.96432205e+00,   6.52230161e-01,   1.07654210e-01,\n",
       "         -1.16280693e-01,   2.31671343e-03,  -1.14737182e-02,\n",
       "          9.66016918e-01,  -1.50723357e-01,   6.40724535e-02,\n",
       "         -1.23561877e-01],\n",
       "       [  5.80318728e-04,   7.82731975e-04,   4.92888418e-02,\n",
       "         -8.63482007e-03,  -5.09278732e-02,   1.76975570e-02,\n",
       "          1.68695302e-01,   1.28888732e-02,   1.10604431e-02,\n",
       "          1.67127112e-02,   1.90648269e-03,  -1.12639677e-03,\n",
       "          7.93408388e-02,  -4.21299848e-02,  -2.50395175e-04,\n",
       "         -5.50917541e-03,   3.59884362e-02,  -6.48330036e-02,\n",
       "          2.59295233e-01,  -7.85881351e-01,   4.40185472e-02,\n",
       "          1.04403245e-01,   3.04859516e-04,  -4.45027286e-03,\n",
       "          1.02223595e-01,   1.48766998e-01,  -8.15299538e-02,\n",
       "          4.17749901e-02],\n",
       "       [ -7.74978313e-04,   4.58263443e-04,   7.34907803e-02,\n",
       "         -1.70283008e-02,   1.66738232e-01,  -1.02823028e-01,\n",
       "         -6.73572566e-01,   2.02399403e-02,   9.73008624e-03,\n",
       "          7.53953047e-02,   3.74778964e-03,   2.95931885e-02,\n",
       "         -1.36704595e-01,   1.80697599e-01,  -9.76678509e-04,\n",
       "          1.07234238e-02,   2.06689602e-02,  -1.99704548e-02,\n",
       "          2.78195276e+00,  -5.33926257e-01,  -1.19193551e-01,\n",
       "         -8.90204157e-02,   1.53312105e-04,   9.68065022e-03,\n",
       "         -3.77175850e-01,   1.14335624e+00,   1.42840307e-01,\n",
       "         -2.24744053e-01],\n",
       "       [ -2.75398866e-04,  -3.56668699e-03,  -5.45388599e-02,\n",
       "         -1.01526814e-02,   4.95822781e-02,  -9.34526525e-02,\n",
       "         -1.87515440e-01,  -1.82059876e-02,  -4.23246828e-03,\n",
       "          2.21715141e-03,  -1.22681453e-02,   8.52212209e-03,\n",
       "          7.72654930e-02,  -1.11320971e-01,   1.42001850e-04,\n",
       "          1.49410540e-03,  -9.52553702e-02,   2.37671156e-02,\n",
       "          9.78885354e-01,  -1.02200764e+00,   7.01725502e-02,\n",
       "          1.23034990e-01,   4.25410554e-03,   2.38914643e-02,\n",
       "          8.59907937e-01,   3.04650469e-01,  -8.41808530e-02,\n",
       "          1.51968881e-01],\n",
       "       [ -3.92835561e-04,   3.61317475e-04,  -1.42853954e-02,\n",
       "         -1.32989053e-04,  -7.68509878e-02,  -4.90468463e-02,\n",
       "          1.58464149e-01,   1.05445710e-02,   4.08595834e-03,\n",
       "          4.60481509e-03,  -2.52109984e-03,   9.27012724e-04,\n",
       "         -3.77500892e-02,   6.82990841e-02,  -1.66170613e-03,\n",
       "         -5.33963425e-03,   5.24283218e-02,  -5.18528573e-03,\n",
       "          7.85174378e-01,  -8.68094830e-01,   3.37243675e-02,\n",
       "          3.02519977e-02,   2.15104058e-03,  -5.09669684e-03,\n",
       "         -3.41809188e-01,   2.75391517e-01,   7.46963655e-02,\n",
       "         -1.33385526e-01],\n",
       "       [ -1.95845631e-03,  -1.42572279e-03,   2.29519872e-01,\n",
       "         -1.14007120e-02,  -1.54288962e-01,   1.55586931e-01,\n",
       "         -4.65993542e-01,  -2.44946501e-02,   6.52753275e-03,\n",
       "         -9.53730557e-02,  -4.11111695e-04,   1.07977124e-02,\n",
       "         -7.17937266e-02,   1.39431934e-01,   3.37692303e-03,\n",
       "         -6.05131352e-03,   2.22820788e-01,   9.30119651e-02,\n",
       "          1.67760365e+00,  -1.85366574e+00,  -7.63907855e-02,\n",
       "         -9.17214515e-02,  -4.64381649e-03,   1.35198709e-02,\n",
       "         -1.02827264e+00,   1.00523180e+00,   1.69336502e-01,\n",
       "         -3.06878127e-01],\n",
       "       [  6.29179986e-04,  -1.08168844e-03,  -1.74535148e-01,\n",
       "         -2.14283628e-03,  -6.36214791e-02,  -1.35932377e-01,\n",
       "         -3.33969199e-01,   5.38846153e-03,   4.43424938e-03,\n",
       "          5.12344625e-02,  -2.12089341e-03,   6.04837678e-03,\n",
       "         -4.31234077e-03,  -6.88738353e-04,  -1.90504143e-03,\n",
       "          6.41786670e-03,   7.67814181e-02,   4.21614774e-02,\n",
       "         -3.76882520e+00,   2.34913670e+00,  -6.41294080e-03,\n",
       "         -3.57927715e-03,   4.08718345e-03,  -4.79497665e-03,\n",
       "          5.43724888e-01,   1.75994421e+00,  -5.81646464e-02,\n",
       "          9.83107600e-02],\n",
       "       [  5.12064292e-04,  -1.06523680e-03,   6.78495945e-02,\n",
       "          7.57352110e-03,   3.89872710e-02,   1.92321537e-01,\n",
       "          6.28240995e-01,   8.05331758e-03,  -3.46093939e-03,\n",
       "         -2.82740699e-02,   1.50690735e-03,   1.61885484e-03,\n",
       "          5.30842788e-02,  -7.11152322e-02,   3.14517328e-04,\n",
       "         -9.98209749e-03,  -9.27973225e-02,  -1.55257688e-03,\n",
       "          3.86369254e+00,  -3.76709007e+00,   2.85549177e-02,\n",
       "         -4.84433883e-03,  -7.83961941e-03,   1.69938240e-02,\n",
       "         -4.10383784e-01,   7.15230890e-01,  -2.55114559e-02,\n",
       "          7.37209847e-03]])"
      ]
     },
     "execution_count": 241,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "coefs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 329,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAY8AAAG4CAYAAAC95WvyAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11172a650>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#objects = ('Reggae','Pop_Rock','Country','Jazz','Vocal','New Age','Latin','Rap','RnB','International','Blues','Electronic','Folk')\n",
    "def plotcoef(coef,labelname):\n",
    "    objects=list(df.columns.values)[1:]\n",
    "    y_pos = np.arange(len(objects))\n",
    "    performance = list(coef)\n",
    "\n",
    "    plt.bar(y_pos, performance, align='center', alpha=0.5,label=labelname)\n",
    "    plt.legend(loc='lower left')\n",
    "    plt.xticks(y_pos, objects,rotation='vertical')\n",
    "    plt.ylabel('Coefficient Values')\n",
    "    \n",
    "plt.title(\"Features Coefficent Comparison for \"+labels[1]+\" and \"+labels[11])\n",
    "plotcoef(coefs[5],labels[5])\n",
    "plotcoef(coefs[2],labels[2])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 275,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.0010822798417954723,\n",
       " -0.0031583077261973873,\n",
       " -0.28807841495570596,\n",
       " 0.015410951394854211,\n",
       " -0.058132451189152654,\n",
       " -0.18445083122641939,\n",
       " 0.67400291296051351,\n",
       " 0.01938804204963205,\n",
       " -0.00839833194943276,\n",
       " 0.00082237171830988175,\n",
       " 0.0010988566905606742,\n",
       " -0.013380895730299077,\n",
       " -0.08344363092766735,\n",
       " 0.16527858313802413,\n",
       " -0.0030583891457569489,\n",
       " -0.0062852925941378653,\n",
       " -0.10008079416993951,\n",
       " 0.097346991196789801,\n",
       " -1.9643220515829762,\n",
       " 0.65223016096513198,\n",
       " 0.10765420983878379,\n",
       " -0.11628069250339348,\n",
       " 0.002316713425601781,\n",
       " -0.011473718170137781,\n",
       " 0.96601691790307709,\n",
       " -0.15072335730592407,\n",
       " 0.064072453469129473,\n",
       " -0.12356187708225405]"
      ]
     },
     "execution_count": 275,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(coefs[5])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}