06initial_analysis.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/albarron/anaconda/lib/python2.7/site-packages/matplotlib/font_manager.py:273: UserWarning: Matplotlib is building the font cache using fc-list. This may take a moment.\n",
      "  warnings.warn('Matplotlib is building the font cache using fc-list. This may take a moment.')\n"
     ]
    }
   ],
   "source": [
    "import sys\n",
    "import re\n",
    "import time\n",
    "import datetime\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import func\n",
    "# inline plot\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(21021, 33)\n"
     ]
    }
   ],
   "source": [
    "df = pd.read_csv(\"data/merged_concat_final.csv\",sep='\\t',error_bad_lines=False)\n",
    "del df['Unnamed: 0']\n",
    "print df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array(['train_id', 'stop_pa', 't_northbound', 'timestamp', 'temp',\n",
       "       'precipiation', 'visability', 'windspeed', 'humidity', 'cloudcover',\n",
       "       'is_bullet', 'is_limited', 'tweet_id', 'is_delay', 'del_min',\n",
       "       'del_med', 'del_maj', 'del_cat', 'is_backlog', 'is_canceled',\n",
       "       'is_passing', 'is_accident', 'is_medical', 'is_mechanical',\n",
       "       'is_customer', 'is_event', 'd_monday', 'd_tuesday', 'd_wednesday',\n",
       "       'd_thursday', 'd_friday', 'd_saturday', 'd_sunday'], dtype=object)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.columns.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "del_min     91\n",
       "del_med    142\n",
       "del_maj     28\n",
       "del_cat     16\n",
       "dtype: int64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[['del_min','del_med','del_maj','del_cat']].sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df.set_index('timestamp')\n",
    "df['timestamp'] = pd.to_datetime(df['timestamp'],format=\"%Y-%m-%d %H:%M:%S\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def delay_to_ordinal(r):\n",
    "    v = 0 #no delay\n",
    "    v = 1 if r['del_min'] == 1 else v\n",
    "    v = 2 if r['del_med'] == 1 else v\n",
    "    v = 3 if r['del_maj'] == 1 else v\n",
    "    v = 4 if r['del_cat'] == 1 else v\n",
    "    return v\n",
    "\n",
    "df['ord_del'] = df.apply(lambda x:delay_to_ordinal(x),axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def days_to_ordinal(r):\n",
    "    v = 0 # Sunday\n",
    "    v = 1 if r['d_monday'] == 1 else v\n",
    "    v = 2 if r['d_tuesday'] == 1 else v\n",
    "    v = 3 if r['d_wednesday'] == 1 else v\n",
    "    v = 4 if r['d_thursday'] == 1 else v\n",
    "    v = 5 if r['d_friday'] == 1 else v\n",
    "    v = 5 if r['d_saturday'] == 1 else v\n",
    "    return v\n",
    "\n",
    "df['ord_weekdays'] = df.apply(lambda x:days_to_ordinal(x),axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "only_delay = df[(df['is_delay']==1)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# df.plot.scatter(x=df['timestamp'],y=df['ord_del'],figsize=[15,6], alpha='0.2')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# only_delay[['del_min','del_med','del_maj','del_cat','is_bullet','is_limited']].plot.hist(color='k',alpha=0.5,stacked=True,bins=4,figsize=[12,6])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "relative to delay\n",
      "is_delay    100.000000\n",
      "del_min      27.828746\n",
      "del_med      43.425076\n",
      "del_maj       8.562691\n",
      "del_cat       4.892966\n",
      "dtype: float64 %\n",
      "Relative to total\n",
      "is_delay    1.555587\n",
      "del_min     0.432900\n",
      "del_med     0.675515\n",
      "del_maj     0.133200\n",
      "del_cat     0.076114\n",
      "dtype: float64 %\n"
     ]
    }
   ],
   "source": [
    "print \"relative to delay\"\n",
    "print (df[['is_delay','del_min','del_med','del_maj','del_cat']].sum()/float(df['is_delay'].sum()))*100 , '%'\n",
    "print \"Relative to total\"\n",
    "print (df[['is_delay','del_min','del_med','del_maj','del_cat']].sum()/float(len(df)))*100 , '%'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Train IDs swapped into cat variables and concat into main dataset\n",
    "train_id_dummies = pd.get_dummies(df['train_id'],prefix='tid')\n",
    "train_id_dummies.shape\n",
    "train_id_dummies.columns.values\n",
    "del train_id_dummies['tid_101.0'] # Delete as base var\n",
    "tid_col = train_id_dummies.columns.values\n",
    "df = pd.concat([df, train_id_dummies], axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Pick one of these to explore re: below models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Look only at train IDs\n",
    "features = df.columns.values\n",
    "X = train_id_dummies\n",
    "y = df['ord_del']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# Non Delay Specific\n",
    "features = df.columns.values\n",
    "target_cols = ['temp','precipiation',\n",
    "        'visability','windspeed','humidity','cloudcover',\n",
    "        'is_bullet','is_limited','t_northbound',\n",
    "       'd_monday','d_tuesday','d_wednesday','d_thursday','d_friday','d_saturday']\n",
    "X = df[target_cols]\n",
    "# del X['is_delay']\n",
    "# del X['tweet_id']\n",
    "# X['timestamp'] = X['timestamp'].apply(lambda x: (np.datetime64(x).astype('uint64') / 1e6).astype('uint32'))\n",
    "# y = df['ord_del']\n",
    "y = df['is_delay']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Including train IDs\n",
    "features = df.columns.values\n",
    "target_cols = ['temp','precipiation',\n",
    "        'visability','windspeed','humidity','cloudcover',\n",
    "        'is_bullet','is_limited','t_northbound',\n",
    "       'd_monday','d_tuesday','d_wednesday','d_thursday','d_friday','d_saturday'] + list(tid_col)\n",
    "X = df[target_cols]\n",
    "# del X['is_delay']\n",
    "# del X['tweet_id']\n",
    "# X['timestamp'] = X['timestamp'].apply(lambda x: (np.datetime64(x).astype('uint64') / 1e6).astype('uint32'))\n",
    "# y = df['ord_del']\n",
    "y = df['is_delay']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# If there IS a delay...\n",
    "features = df.columns.values\n",
    "X = only_delay[['is_backlog', 'is_canceled',\n",
    "       'is_passing', 'is_accident', 'is_medical', 'is_mechanical',\n",
    "       'is_customer', 'is_event']]\n",
    "# del X['is_delay']\n",
    "# del X['tweet_id']\n",
    "# X['timestamp'] = X['timestamp'].apply(lambda x: (np.datetime64(x).astype('uint64') / 1e6).astype('uint32'))\n",
    "y = df['ord_del']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/albarron/anaconda/lib/python2.7/site-packages/ipykernel/__main__.py:4: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n"
     ]
    }
   ],
   "source": [
    "# X['timestamp'] = X['timestamp'].apply(lambda x:int(x))\n",
    "# X['stop_pa'] = X['stop_pa'].apply(lambda x:int(x))\n",
    "# X['train_id'] = X['train_id'].apply(lambda x:int(x))\n",
    "X['t_northbound'] = X['t_northbound'].apply(lambda x:int(x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/albarron/anaconda/lib/python2.7/site-packages/ipykernel/__main__.py:1: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  if __name__ == '__main__':\n"
     ]
    }
   ],
   "source": [
    "X['cloudcover'] = X['cloudcover'].fillna(X['cloudcover'].mean())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# X.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# df.plot.scatter(x='timestamp',y='del_ord',figsize=[15,5])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "X_y = only_delay[['is_delay','ord_del','temp','precipiation',\n",
    "        'visability','windspeed','humidity','cloudcover',\n",
    "        'is_bullet','is_limited','t_northbound',\n",
    "       'd_monday','d_tuesday','d_wednesday','d_thursday','d_friday','d_saturday']]\n",
    "cr = X_y.corr()\n",
    "np.round(cr, 4)\n",
    "# "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "X_y.sum()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Run Decision Trees, Prune, and consider False Positives"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DecisionTreeClassifier(class_weight=None, criterion='gini', max_depth=2,\n",
       "            max_features=None, max_leaf_nodes=None, min_samples_leaf=5,\n",
       "            min_samples_split=2, min_weight_fraction_leaf=0.0,\n",
       "            random_state=None, splitter='best')"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.tree import DecisionTreeClassifier\n",
    "TreeClass = DecisionTreeClassifier(\n",
    "                max_depth = 2,\n",
    "                min_samples_leaf = 5)\n",
    "TreeClass.fit(X,y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.992055614405\n"
     ]
    }
   ],
   "source": [
    "from sklearn.cross_validation import cross_val_score\n",
    "scores = cross_val_score(TreeClass, X, y, cv=10)\n",
    "print(scores.mean()) # Score = More is better, error is 1-score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[20648    46]\n",
      " [  121   206]]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import confusion_matrix\n",
    "y_hat = TreeClass.predict(X)\n",
    "cmat = confusion_matrix(y, y_hat)\n",
    "print cmat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 0.00582599  0.00582599  0.00582599 ...,  0.00582599  0.00582599\n",
      "  0.00582599]\n",
      "0.813990833077\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import roc_curve, auc,roc_auc_score\n",
    "y_hat_probability = TreeClass.predict_proba(X).T[1] \n",
    "print(y_hat_probability)\n",
    "print(roc_auc_score(y, y_hat_probability))\n",
    "vals = roc_curve(y, y_hat_probability)  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x10fff6310>"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXkAAAEQCAYAAABV+ASvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzt3Xt0VOW9//H3l5sihJSLoCAEBSkiCKJyUS5jtYAuWyy1\nliBYpMdSqR79/WrVrlq1eqx2tVqP14pVrP2J9ngqlIJ3FwHkrogKgihqgAAigibKRS7f3x97T2YS\nSTIJk5lk5/NaaxazZ57Z+5u9hk92nr3385i7IyIi0dQo2wWIiEjtUciLiESYQl5EJMIU8iIiEaaQ\nFxGJMIW8iEiEVRnyZvaomX1iZm9X0uZeM3vfzFaaWb/0ligiIjWVypH8NGBkRW+a2XlAN3c/EZgM\n/CVNtYmIyGGqMuTd/TVgZyVNRgNPhG2XArlm1iE95YmIyOFIR598J2Bj0nJR+JqIiGSZTryKiERY\nkzSsowjonLR8XPjaN5iZBsoREakBd7eafC7VI3kLH4cyC7gUwMwGAZ+7+ycVrcjd9XDn5ptvznoN\ndeWhfaF90dD2xd69zsKFzu23OyNGODk5Tt++zn/+p/Pss86nn5ZtfziqPJI3s+lADGhrZhuAm4Fm\nQV77VHd/zszON7MPgK+Ayw6rIhGRiNm7F5Ytg4ICmDcPli6FHj1g+HCYMgWeegratKmdbVcZ8u4+\nLoU2V6anHBGR+m/PHliyJAj0ggJ4/XU46aQg1K+5BoYMgW99KzO1pKNPXmogFotlu4Q6Q/siQfsi\noT7ti127YPHiRKivWAG9e0MsBtddB2edBa1aZac2O9z+nmptzMwzuT0Rkdrw5ZewaFEQ6vPmwcqV\n0LdvcKQei8GZZ0LLlunbnpnhNTzxWidCvmvXrhQWFmasDpG8vDw+/vjjbJch9URxMSxcmAj1d96B\n/v0ToT54MBx1VO1tv96HfPgDZKwOEX3npDJffAELFiRCfc0aOP30INSHD4dBg6B588zVo5AXqSZ9\n5yTZzp1BqMevflm3DgYOTIT6gAFw5JHZq08hL1JN+s41bJ99BvPnJ06UfvhhcHQeiwWhfsYZ0KxZ\ntqtMUMiLVJO+cw3Ltm1lQ33DhuDkaDzUTzsNmjbNdpUVU8iLVJO+c9G2dWuiP33ePCgqCq5Nj58o\nPfVUaFKPLiBXyEuNjBw5kokTJ5Kfn3/I9y+//HK6devGDTfckOHKap++c9FSVFQ21Ldtg6FDE33q\n/fpB48bZrrLmDifkMzpeQ7C5b6ro9bqgZcuWnpOT4zk5Od6oUSNv3rx56WvTp0/PWB033nijN23a\n1HNycrx169Y+ZMgQX7p0adrW/9e//tVjsVja1peKs846y4888kjPycnx9u3b+0UXXeTbtm1L6bP7\n9+93M/PCwsIabbsuf+ekahs2uD/xhPtPf+revbt727buP/iB+z33uL/5pvv+/dmuML3C72uNcldD\nDVehpKSE4uJiiouLycvLY86cOaWvHeoI+MCBA7VWy/jx4ykuLmbbtm0MGDCAH/7wh2lbt7tjVrMD\nhZoyM6ZOnUpxcTHr1q1j586dXHfddSl9Nhv1SvZ8/DH87W9w2WVwwglBH/qsWcER+rPPBkfuzz4L\nV19d/4/a000hXw2e+Iuk1G9/+1vGjh3LuHHjyM3N5cknn2TChAnceuutpW1effVVjj/++NLloqIi\nxowZQ/v27enWrRsPPvhgtepo0qQJP/nJT9i8eTPFxcW4O7feeitdu3blmGOOYdKkSZSUlACwe/du\nLrnkEtq1a0fr1q0ZNGgQO3bsAGDo0KE88cQTrFq1iquuuooFCxaQk5ND+/btAcr8HD169OCll14q\nrWHfvn20bduWVatWAbBw4UIGDx5M69at6d+/PwsWLEh5nwLk5uYyevRoVq5cWfrekiVLStfZqVMn\nrr766tJfosOHDwegV69etGrVihkzZgAwa9Ys+vXrR+vWrRk2bBirV6+u1r6V7HOH9evhscfgJz+B\nvLzgypfnngsuZZw9Gz75BJ55Bq68Evr0gUZKsgrVm11jdviP2jJz5kzGjx/PF198wcUXX1xB/UEB\n7s4FF1zAwIED2bJlCy+//DJ/+tOfmDt3bsrb27t3L9OmTaNr1660atWKRx55hOnTpzN//nzWr1/P\njh07uOaaawCYNm0au3fvZvPmzezYsYMHH3yQI8td8Nu7d2/uv/9+hg4dSklJCdu2bfvGNvPz85k+\nfXrp8nPPPUenTp3o3bs3GzduZPTo0dx2223s3LmTO++8kzFjxrBzZ2WzRpa1fft2ZsyYwYknnlj6\nWtOmTbn33nvZsWMHCxcu5MUXX+Thhx8GYP78+bg7a9asobi4mB/84AcsX76cyZMn89hjj7Fjxw4m\nTZrE6NGj2b9/f8p1SOa5B9elP/IIjB8PXbrAsGHwyivBmC8vvQRbtsA//gFXXAG9etXu/+eoqTch\n7374j9oyZMgQzj//fIBvBGh5ixYtoqSkhOuvv57GjRtzwgknMGnSJJ5++ukqt/Pkk0/Spk0b8vLy\nWL16NTNnzgRg+vTpXHvttXTp0oUWLVrw+9//vjSQmzZtyvbt21m3bh1mRv/+/TmqBvdfjxs3jpkz\nZ/L1118D8NRTTzFuXDBA6d///ndGjx7NueeeC8CIESPo27cvL7zwQpXrnTJlCq1bt6Z9+/aUlJRw\nzz33lL532mmnccYZZ2BmdO3alcsvv5x58+aV+XzyX1aPPPIIU6ZMoX///pgZEydOBGD58uXV/nml\n9rjD2rXwl79Afj506gTnnhtc4hiLwdy5sGkTTJ8OP/sZfPvbCvXDUY8uIqq7OnfuXHWj0IYNGygs\nLKRNOHi0u3Pw4EHOPvvsKj97ySWX8Nhjj33j9c2bN5OXl1e6nJeXx969e/n000+ZOHEiW7Zs4eKL\nL6akpITx48dz++2306iaf99++9vfpnv37syZM4cRI0Ywe/Zs/vjHPwJQWFjI9OnTS7tM3J39+/dz\n3nnnVbnehx56iAkTJvDOO+/wve99j6KiIjp27AjAe++9xy9/+UveeOMNdu3axYEDBxg4cGCF64rX\n8ec//7m0jn379lFUdMiJyiRD3OHddxN3k86fHwwJMHw4jBgBv/89dO2qIK8tCvk0KH8CsEWLFuza\ntat0ecuWLaXPO3fuTI8ePdLaV9yxY8cyA7wVFhZyxBFHcPTRRwNw0003cdNNN1FYWMjIkSPp1asX\nEyZMqPRnOJSxY8cyffp0vvrqK0499dTSX26dO3dm0qRJPPDAA9WuPX4k3qdPH2644QZ+8YtfsGzZ\nMgAmT57M4MGDeeaZZ2jevDl33XUXc+bMqbDezp07c/PNN/OrX/2q2nVI+hw8CKtWJW48mj8/GGY3\nFoMLLoA//jHoZ5fMqDfdNfVJv379mDNnDp9//jlbtmzhvvvuK31v8ODBNGvWjLvvvpu9e/dy4MAB\nVq1axYoVK2q8vfz8fO6++24KCwspKSnhxhtvLO1KmTt3LqtXr8bdadmyJU2bNqXxIS496NChA5s2\nbaq0/zo/P5/nn3+eqVOnlq4fghO0M2bM4JVXXuHgwYPs2bOHgoICtm7dWq2fY9KkSWzcuJHnn38e\nCK5sys3NpXnz5qxZs6a0Px6gUaNGtGvXjg8//LD0tcsvv5wHHniA119/HYAvv/yS2bNns3v37mrV\nIdVz4AC8+Sbccw9ceCEcfTRcdFEwUuOYMcF769fDo4/CpZcq4DNNIV8NqV6yN3HiRHr27EleXh7n\nn39+mUstGzduzHPPPceyZcvo2rUr7du35+c//3np1TA1cfnll/PjH/+YoUOH0r17d3Jzc0v7tjdv\n3syYMWPIzc2lT58+jBgxorSe5J/nu9/9LieeeCIdOnQo7S4pr1OnTpx++uksW7aszAnmvLw8ZsyY\nwW233cbRRx9N165dufvuuzl48GCldZffn82aNeOqq67itttuA+Cuu+7i8ccfp1WrVlxxxRWMHTu2\nTPvf/e535Ofn06ZNG2bOnMnAgQN56KGHuOKKK2jTpg09e/bkySefTHEvSqr27w9mOrrrLvj+96Fd\nOxg3LuhnHzs2CPd162DqVLjkEjjuuGxX3LDpjldpkPSdS93+/fDGG4m7SRcuDII7PkTAsGHQoUO2\nq4w2DWsgUk36zlVs377gSD1+onTx4uDEaHyIgGHDgi4ZyRyFfESMHDmSRYsWlbmm3sy46aabuPba\na7NcXfUdPHiQ3NzcMt0y8Z/p5ZdfrvRKmdqm71zC3r2wfHki1Jcuhe7dE6E+dCi0bZvtKhs2hbxI\nNTXk79yePUGQx69+Wb48uBY9PuzukCHQunW2q5RkCnmRampI37ndu4Mul3iov/EGnHxyok/9rLMg\nNzfbVUplFPIi1RTl79xXX8GiRYkTpW++Caeckgj1M8+EnJxsVynVcTghXyduhsrLy9OIgpJReRG6\nWLukJLjiJR7qb78dTIoxfDjcdFMQ6i1aZLtKyZY6cSQvIqkrLobXXkucKF29Gk4/PXGidNAgqMHw\nRFKH1fvuGhGp2Oefw4IFiT71994LhtyNh/rAgVDFuHhSzynkRSJkx46yk05/8EFwdB6/+uWMM+CI\nI7JdpWSSQl6kHvv000Soz5sHH30U9KPHT5Sedho0a5btKiWbFPIi9cgnn5SddHrTpuAyxnio9+8P\nTerEJRFSVyjkReqwzZvLhvrWrcFdpPE+9X79FOpSOYW8SB2yaVOiP33ePPjss2C8l3ion3KKJpqW\n6lHIi2RRYWHiKL2gILjEMR7ow4dD796aaFoOj0JeJEPcgxOjyaG+e3eiP334cDjpJIW6pJdCXqSW\nuAezGsW7XubNC8ZXTw51TTQttU0hL5Im7sGsRsmh3qhRouslFguG4VWoSybVesib2SjgHoLpAh91\n9z+Ue78V8P+ALkBj4C53f/wQ61HIS53iDmvWlD1ResQRiaP0WAyOP16hLtlVqyFvZo2AdcA5wGZg\nOTDW3dcmtfk10Mrdf21m7YD3gA7uvr/cuhTyklUHDwZjvcRDff58aNkyEerDhwezIInUJbU9CuUA\n4H13Lww39jQwGlib1MaB+OClOcBn5QNeJBsOHgxGZYx3vcyfH0yIMXw4XHgh/PnP0LlztqsUqT2p\nhHwnYGPS8iaC4E92PzDLzDYDLYEfp6c8keo5cABWrkyE+oIF0L59EOo/+hHcdx906pTtKkUyJ133\n2Y0E3nT375hZN+BlMzvF3b8s3/CWW24pfR6LxYjFYmkqQRqi/fthxYpEqC9cCB07BqE+bhw8/DAc\nc0y2qxSpnoKCAgoKCtKyrlT65AcBt7j7qHD5BsCTT76a2WzgDndfGC6/Clzv7q+XW5f65OWw7NsX\nTF8XP0m6aBHk5SX604cNC47cRaKktvvklwPdzSwP2AKMBfLLtSkEzgUWmlkHoAfwYU0KEkn29dfB\nRNPxE6VLlkC3bkGg/+xn8Pe/Q7t22a5SpO6qziWU/03iEso7zWwywRH9VDM7FngcODb8yB3u/tQh\n1qMjeanU3r2wdGki1Jctgx49Ele/DB0anDgVaUh0M5TUW7t3B0fn8T715cuhV6/ENepDhkBubrar\nFMkuhbzUG7t2Bf3o8VBfsQL69En0qZ91FrRqle0qReoWhbzUWV9+GYR6/ETpW28F46fHQ/3MM4Ob\nkUSkYgp5qTOKi4PLGOOhvmpVMH1dPNQHD4ajjsp2lSL1i0Jesubzz+G11xInStesCSaajp8oHTgQ\nmjfPdpUi9ZtCXjJmx47gLtJ4n/q6dUGQx0N9wIBggC8RSR+FvNSa7duD8V7iof7hh0GXS/zql9NP\nh2bNsl2lSLQp5CVttm0rO+n0hg3BFS/xUO/fH5o2zXaVIg2LQl5qbOvWsmOpb94c3HAUP1F66qnQ\nJF0jHIlIjSjkJWVFRWXnJ/3002C8l3io9+0LjRtnu0oRSaaQlwpt2FA21D//PAj1+InSPn006bRI\nXaeQl1Iff1x2ftIvvyw76XSvXgp1kfpGId9AuQdXuySH+tdfl510umdPzU8qUt8p5BsId3j//bKh\nDmVD/cQTFeoiUaOQjyh3WLu27NUvTZuWnXS6WzeFukjUKeQj4uBBePfdRKjPnx+M85Ic6l27KtRF\nGhqFfD118CC8806i62X+/GDs9OQTpV26ZLtKEck2hXw9ceBAMNRuPNQXLAimrosfpQ8fDscdl+0q\nRaSuUcjXUfv3w8qVif70116DY48tG+rHHlvlakSkgVPI1xH79gUzHcX71Bctgs6dE4E+bBh06JDt\nKkWkvlHIZ8nXX8PrrydCffFiOP74RH/6sGFBd4yIyOFQyGfI3r2wbFmiT33JkuC69HioDx0Kbdpk\nu0oRiRqFfC3ZsycI8nioL18e3EEav/plyBD41reyXaWIRJ1CPk127QpCPX6i9I03oHfvRJ/6kCHQ\nqlW2qxSRhkYhX0NffRWcHI2H+sqVwVC78VA/80zIycl2lSLS0CnkU1RSAgsXJk6UvvNOMNNRPNQH\nD4YWLbJWnojIISnkK/DFF8G16fFQf/fdYE7S+InSQYOgefOMlSMiUiMK+dDOncFdpPETpe+9BwMG\nJEJ9wAA48sha27yISK1osCH/2WfBeC/xUF+/Pjg6j1/9csYZ0KxZ2jYnIpIVDSbkP/00CPX4idLC\nwuDkaDzUTzstGIpXRCRKIhvyn3xSdn7SoqLgMsb4idL+/aFJk9qrV0SkLohMyG/eXDbUt20LQj3e\np96vHzRunLFyRUTqhHob8hs3lg31HTuC8V7iod6nj0JdRKRehfzjj3tpsJeUBKEe71M/+WRo1Chj\n5YiI1Av1KuQHDnQuvTQI9ZNO0lR2IiJVOZyQz/hpy/x8mDIl01sVEWmYUuocMbNRZrbWzNaZ2fUV\ntImZ2ZtmtsrM5la8rpqWKiIi1VXlkbyZNQLuB84BNgPLzexf7r42qU0u8AAwwt2LzKzCqTLU5y4i\nkjmpRO4A4H13L3T3fcDTwOhybcYB/3T3IgB3317hBhXyIiIZk0rkdgI2Ji1vCl9L1gNoY2ZzzWy5\nmU2ocIMKeRGRjEnXidcmQH/gO0ALYLGZLXb3D8o3VJ+8iEjmpBLyRUCXpOXjwteSbQK2u/seYI+Z\nzQf6At8I+dmzb2HLluB5LBYjFotVv2oRkQgrKCigoKAgLeuq8jp5M2sMvEdw4nULsAzId/c1SW16\nAvcBo4AjgKXAj9393XLr8r/+1fnpT9NSu4hIg1Cr18m7+wEzuxJ4iaAP/1F3X2Nmk4O3faq7rzWz\nF4G3gQPA1PIBnyi2JmWKiEhNZPyO12nTnIkTM7ZJEZF673CO5DN+rYuurhERyZyMR666a0REMkdH\n8iIiEaaQFxGJMIW8iEiEqU9eRCTCdCQvIhJhCnkRkQhTd42ISITpSF5EJMIU8iIiEabuGhGRCNOR\nvIhIhCnkRUQiTCEvIhJh6pMXEYkwHcmLiESYQl5EJMLUXSMiEmE6khcRiTCFvIhIhCnkRUQiTH3y\nIiIRpiN5EZEIU8iLiESYumtERCJMR/IiIhGmkBcRiTB114iIRJiO5EVEIkwhLyISYQp5EZEIU5+8\niEiE6UheRCTCFPIiIhGm7hoRkQhLKeTNbJSZrTWzdWZ2fSXtzjCzfWY2psIN6kheRCRjqoxcM2sE\n3A+MBE4G8s2sZwXt7gRerHSDCnkRkYxJJXIHAO+7e6G77wOeBkYfot1VwP8C2ypbmbprREQyJ5WQ\n7wRsTFreFL5Wysw6Ahe6+0NApTGuI3kRkcxJV+TeAyT31VcY9Ap5EZHMaZJCmyKgS9LyceFryU4H\nnjYzA9oB55nZPnefVX5ld999Cy1bBs9jsRixWKwGZYuIRFdBQQEFBQVpWZe5e+UNzBoD7wHnAFuA\nZUC+u6+poP004N/u/uwh3vOtW50OHQ67bhGRBsPMcPcandGs8kje3Q+Y2ZXASwTdO4+6+xozmxy8\n7VPLf6Sy9am7RkQkc6o8kk/rxsx8+3anbduMbVJEpN47nCN53fEqIhJhGrtGRCTCFPIiIhGm7hoR\nkQjTkbyISIQp5EVEIkwhLyISYeqTFxGJMB3Ji4hEmEJeRCTC1F0jIhJhCnkRkQhT54mISIQp5EVE\nIkwhLyISYQp5EZEIU8iLiESYQl5EJMIU8iIiEaaQFxGJMIW8iEiEKeRFRCJMIS8iEmEKeRGRCFPI\ni4hEmEJeRCTCFPIiIhGmkBcRiTCFvIhIhCnkRUQiTCEvIhJhCnkRkQhTyIuIRJhCXkQkwhTyIiIR\nppAXEYmwlELezEaZ2VozW2dm1x/i/XFm9lb4eM3M+qS/VBERqS5z98obmDUC1gHnAJuB5cBYd1+b\n1GYQsMbdvzCzUcAt7j7oEOvyqrYnIiJlmRnubjX5bCpH8gOA99290N33AU8Do5MbuPsSd/8iXFwC\ndKpJMSIikl6phHwnYGPS8iYqD/H/AJ4/nKJERCQ9mqRzZWZ2NnAZMKSiNrfcckvp81gsRiwWS2cJ\nIiL1XkFBAQUFBWlZVyp98oMI+thHhcs3AO7ufyjX7hTgn8Aod19fwbrUJy8iUk213Se/HOhuZnlm\n1gwYC8wqV0AXgoCfUFHAi4hI5lXZXePuB8zsSuAlgl8Kj7r7GjObHLztU4HfAm2AB83MgH3uPqA2\nCxcRkapV2V2T1o2pu0ZEpNpqu7tGRETqKYW8iEiEKeRFRCJMIS8iEmEKeRGRCFPIi4hEmEJeRCTC\nFPIiIhGmkBcRiTCFvIhIhCnkRUQiTCEvIhJhCnkRkQhTyIuIRJhCXkQkwhTyIiIRppAXEYkwhbyI\nSIQp5EVEIkwhLyISYQp5EZEIU8iLiESYQl5EJMIU8iIiEaaQFxGJMIW8iEiEKeRFRCJMIS8iEmEK\neRGRCFPIi4hEmEJeRCTCFPIiIhGmkBcRiTCFvIhIhCnkRUQiLKWQN7NRZrbWzNaZ2fUVtLnXzN43\ns5Vm1i+9ZYqISE1UGfJm1gi4HxgJnAzkm1nPcm3OA7q5+4nAZOAvtVBrpBQUFGS7hDpD+yJB+yJB\n+yI9UjmSHwC87+6F7r4PeBoYXa7NaOAJAHdfCuSaWYe0Vhox+gInaF8kaF8kaF+kRyoh3wnYmLS8\nKXytsjZFh2gjIiIZphOvIiIRZu5eeQOzQcAt7j4qXL4BcHf/Q1KbvwBz3f0f4fJaYLi7f1JuXZVv\nTEREDsndrSafa5JCm+VAdzPLA7YAY4H8cm1mAb8A/hH+Uvi8fMAfTpEiIlIzVYa8ux8wsyuBlwi6\ndx519zVmNjl426e6+3Nmdr6ZfQB8BVxWu2WLiEgqquyuERGR+qtWTrzq5qmEqvaFmY0zs7fCx2tm\n1icbdWZCKt+LsN0ZZrbPzMZksr5MSvH/SMzM3jSzVWY2N9M1ZkoK/0damdmsMCveMbOJWSiz1pnZ\no2b2iZm9XUmb6uemu6f1QfCL4wMgD2gKrAR6lmtzHjAnfD4QWJLuOurCI8V9MQjIDZ+Pasj7Iqnd\nq8BsYEy2687i9yIXWA10CpfbZbvuLO6LXwN3xPcD8BnQJNu118K+GAL0A96u4P0a5WZtHMnr5qmE\nKveFuy9x9y/CxSVE9/6CVL4XAFcB/wtsy2RxGZbKvhgH/NPdiwDcfXuGa8yUVPaFAznh8xzgM3ff\nn8EaM8LdXwN2VtKkRrlZGyGvm6cSUtkXyf4DeL5WK8qeKveFmXUELnT3h4AoX4mVyveiB9DGzOaa\n2XIzm5Cx6jIrlX1xP9DLzDYDbwFXZ6i2uqZGuZnKJZSSAWZ2NsFVSUOyXUsW3QMk98lGOeir0gTo\nD3wHaAEsNrPF7v5BdsvKipHAm+7+HTPrBrxsZqe4+5fZLqw+qI2QLwK6JC0fF75Wvk3nKtpEQSr7\nAjM7BZgKjHL3yv5cq89S2RenA0+bmRH0vZ5nZvvcfVaGasyUVPbFJmC7u+8B9pjZfKAvQf91lKSy\nLy4D7gBw9/Vm9hHQE3g9IxXWHTXKzdrorim9ecrMmhHcPFX+P+ks4FIovaP2kDdPRUCV+8LMugD/\nBCa4+/os1JgpVe4Ldz8hfBxP0C8/JYIBD6n9H/kXMMTMGpvZUQQn2tZkuM5MSGVfFALnAoR90D2A\nDzNaZeYYFf8FW6PcTPuRvOvmqVKp7Avgt0Ab4MHwCHafuw/IXtW1I8V9UeYjGS8yQ1L8P7LWzF4E\n3gYOAFPd/d0sll0rUvxe/BfweNKlhde5+44slVxrzGw6EAPamtkG4GagGYeZm7oZSkQkwjQKpYhI\nhCnkRUQiTCEvIhJhCnkRkQhTyIuIRJhCXkQkwhTyIiIRppCXtDGzA2a2IhwDfUV4N29FbfPM7J00\nbvvjpHH5XzCz9jVYx+/M7Dvh86vN7Mik92abWas01hvfV++Y2b+qWreZ5ZrZFenavjQcuhlK0sbM\nit09pSC0YM7gf7v7KWna9ofAae6+08xuB1q4+zWHsb6PwvXVyp2VyfvKzB4H3nP3Oypp35Vgf0V2\nUhmpHTqSl3T6xpgb4RH7fDN7PXwMOkSbXma2NDyyXRmONIiZXZL0+kPhsA+VbTv+/nyge7iOfDN7\nO3zcGb7WyMymha+9ZWZXh69PM7MxZnYV0BGYa2avhu99ZGZtzOwOM5uSVPvNZvZ/w+fXmtmy8Ge4\nuRr7bTHhkLFm1sLMXgn31Vtm9r2wzR3ACeG++MNhbk8akmzPhqJHdB7AfmAF8CbBhBcARwLNwufd\ngeXh8zzCGXCAe4H88HkT4AiCUQZnAY3D1x8Axley7Y+ANuHz+whC8ViCwa3akJhx6vsEQ/i+lPTZ\nVuG/0whnowrX1zqpzYfhevoBBUmvryYI6O8CD4evGfBvYEgl9ZaE/zYG/gcYES43AlqGz9sSTKhR\nZn+Fy9Xanh4N96Hx5CWddrl7/3KvNQPut2A+ygPAiYf43GLgN2bWGXjW3T8ws3MIwnh5eAR/JFDV\niHtzzewAwaBevyEY7Gmuh10uZvYkMIxgwKvjzey/gecIBsc6FCv/3N1XmtnRZnYM0B7Y4e5FZnYN\n8F0zWxG2bRH+rK9VsO7mYdvjgHeBl8PXGwF3mNkw4CDQsYLzCyOquT1poBTyUtv+D7DV3U8xs8bA\n7vIN3P0ZPlARAAAB0ElEQVQpM1sCXADMCUcgNOBv7v6bamwr5knj8Ye9O9/o4nH3z82sL8FkFD8H\nfkQwK1eqngk/cwzwj/jmCOYhfSTFdexy9/7hyd0XgV8QzIB0CcFY+qe6+8Hw3MCRh/h8dbcnDZT6\n5CWdDtVnngtsCZ9fStA9UfZDZse7+0fufh9BF80pBF0rF5nZ0WGb1pVdrVPB9pcBw8K+9MZAPjDP\nzNoSdAPNAG4k+IuhvGKgopPI/0Mw7vkPCQIfgqCeZGYtwno7xmuvrFYPJgW5GrjWzBoR7K9tYcCf\nTdBNA1BCYp7TmmxPGigdyUs6HepSrQeBf5rZpcALBONgl3exBXOY7iP4hXB7eLR9I/BSGH5fExzt\nbkh12+6+1cxuAArCl2a7+78tmIlrWrheB244xDoeAV4wsyJ3Pyf5PXd/18xygE0eTtrg7i+bWU+C\nafogCOXxwKdV1Rt2Ab1F8EvoSeDf4fLrhBOFuPsOM1towZjqz7v79WZ2UjW2Jw2ULqEUEYkwddeI\niESYumukXglP0DaLLxJ0e0xw99XZq+rQzKwNwbmF+J/L8XrP8ehO2C51jLprREQiTN01IiIRppAX\nEYkwhbyISIQp5EVEIkwhLyISYf8fdRePzueZn8MAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x115f78e10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "Roc_DataFrame = pd.DataFrame({'False_Positive_Rate':vals[0],'True_Positive_Rate':vals[1]})\n",
    "Roc_DataFrame.plot(x = 'False_Positive_Rate' , y = 'True_Positive_Rate' ) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### As a check, consider Feature selection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('d_friday', 0.36280000000000001),\n",
       " ('windspeed', 0.34889999999999999),\n",
       " ('t_northbound', 0.15679999999999999),\n",
       " ('precipiation', 0.1358),\n",
       " ('humidity', 0.048800000000000003),\n",
       " ('cloudcover', 0.0465),\n",
       " ('d_saturday', 0.0047999999999999996),\n",
       " ('visability', 0.0016000000000000001),\n",
       " ('d_monday', 0.001),\n",
       " ('d_tuesday', 0.0001),\n",
       " ('temp', 0.0),\n",
       " ('is_bullet', 0.0),\n",
       " ('is_limited', 0.0),\n",
       " ('d_wednesday', 0.0),\n",
       " ('d_thursday', 0.0)]"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn import feature_selection\n",
    "pvals = feature_selection.f_regression(X,y)[1] \n",
    "sorted(zip(X.columns.values,np.round(pvals,4)),key=lambda x:x[1],reverse=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x110010290>"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEWCAYAAABMoxE0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n8v8BIBImjuPxtWoAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10ff9c710>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "X_lr=df[['windspeed','t_northbound','precipiation','d_friday']]\n",
    "# localize your search around the maximum value you found\n",
    "c_list = np.logspace(-1,1,21) \n",
    "c_index = np.linspace(-1,1,21)\n",
    "#C is just the inverse of Lambda - the smaller the C - the stronger the\n",
    "#regulatization. The smaller C's choose less variables\n",
    "cv_scores = []\n",
    "for c_score in c_list:\n",
    "    lm = LogisticRegression(C = c_score, penalty = \"l1\")\n",
    "    cv_scores.append(cross_val_score(lm,X,y,cv=10).mean())\n",
    "\n",
    "\n",
    "C_Choice_df = pd.DataFrame({'cv_scores': cv_scores ,'Log_C': c_index })\n",
    "C_Choice_df.plot(x ='Log_C',y = 'cv_scores' )\n",
    "# it sounds like our best choice is C = -0.1  (we chose the most restrictive option)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Find the Principal Components"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "X = only_delay[['temp','precipiation',\n",
    "        'visability','windspeed','humidity','cloudcover',\n",
    "        'is_bullet','is_limited','t_northbound',\n",
    "       'd_monday','d_tuesday','d_wednesday','d_thursday','d_friday','d_saturday']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(21021, 4)"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.decomposition import PCA\n",
    "clf = PCA(.99)\n",
    "X_trans = clf.fit_transform(X)\n",
    "X_trans.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Exp Var ratio: [ 0.94377417  0.03863902  0.00479571  0.00322025]\n",
      "PCA Score: -7.97024397141\n"
     ]
    }
   ],
   "source": [
    "print \"Exp Var ratio:\",clf.explained_variance_ratio_\n",
    "print \"PCA Score:\",clf.score(X,y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAVwAAAEACAYAAAD/Ws6gAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ni8USElbgWiwWS0j8/4dmIEQYP/xKAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10de2a8d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "plt.scatter(X_trans[:, 0], X_trans[:, 1],c=y, alpha=0.2)\n",
    "plt.colorbar();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "collapsed": false
   },
   "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",
       "          penalty='l2', random_state=None, solver='liblinear', tol=0.0001,\n",
       "          verbose=0)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "lm = LogisticRegression()\n",
    "lm.fit(X_trans,y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[-4.22049215]\n",
      "[[-0.04419672 -0.04835333  0.31824815 -0.15575732]]\n"
     ]
    }
   ],
   "source": [
    "print(lm.intercept_)\n",
    "print(lm.coef_)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.984444199569\n",
      "0.0155558004307\n"
     ]
    }
   ],
   "source": [
    "from sklearn.cross_validation import cross_val_score\n",
    "print(cross_val_score(lm,X_trans,y,cv=10).mean()) \n",
    "MisClassificationError = 1 - (cross_val_score(lm,X_trans,y,cv=10).mean())\n",
    "print(MisClassificationError)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Seeing if I can get anything interesting out of KNN given above"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Lecture 10, look at Confusion matrix and ROC curve. Fiddle with the thresholds and AUC"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "12.5654166667\n",
      "2.63625\n"
     ]
    }
   ],
   "source": [
    "print df['windspeed'].max()\n",
    "print df['windspeed'].min()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "df['windspeed_st'] = df['windspeed'].apply(lambda x:x/15.0) # Ballparking "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "\n",
    "X_reg = df[['precipiation','d_friday','t_northbound','windspeed_st']]\n",
    "y_reg = df['is_delay']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x10fc06990>"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3QAAAFHCAYAAADp4Q6RAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nnSRJkiRVqf8f0jga7FrrdggAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x116899ed0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn import cross_validation\n",
    "from sklearn import neighbors, metrics\n",
    "kf = cross_validation.KFold(len(X_reg), n_folds = 10, shuffle = True) #10 fold CV\n",
    "Score_KNN_CV = []\n",
    "RangeOfK = range(1,20) \n",
    "scores = []\n",
    "for k in RangeOfK:\n",
    "    knn = neighbors.KNeighborsClassifier(n_neighbors=k, weights='uniform')\n",
    "    scores = []\n",
    "    for train_index, test_index in kf:        \n",
    "        knn.fit(X_reg.iloc[train_index], y_reg.iloc[train_index])\n",
    "        scores.append(knn.score(X_reg.iloc[test_index],y_reg.iloc[test_index]))\n",
    "    Score_KNN_CV.append(np.mean(scores))\n",
    "\n",
    "Score_KNN_CV_df = pd.DataFrame({'Score_KNN_CV': Score_KNN_CV ,'K': RangeOfK })\n",
    "Score_KNN_CV_df.plot(x = 'K',y = 'Score_KNN_CV',figsize=[15,5])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Cross Validation and Random Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.cross_validation import cross_val_score\n",
    "RFClass = RandomForestClassifier(n_estimators = 10000, \n",
    "                                 max_features = 4, # You can set it to a number or 'sqrt', 'log2', etc\n",
    "                                 min_samples_leaf = 5,\n",
    "                                 oob_score = True)\n",
    "RFClass.fit(X,y)\n",
    "print(RFClass.oob_score_)\n",
    "scores = cross_val_score(RFClass, X, y, cv=10)\n",
    "print(scores.mean())\n",
    "#out of bag error = 25% , CV_error is 35%"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "RFClass.fit(X,y)\n",
    "ImportanceDataFrame = pd.DataFrame({'feature':X.columns.values, 'importance':RFClass.feature_importances_})\n",
    "ImportanceDataFrame.sort_values(by = ['importance'],ascending = 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "Depth_Choice_df = pd.DataFrame({'cv_scores': score,'Number of Features': Features})\n",
    "Depth_Choice_df.plot(x ='Number of Features',y = 'cv_scores')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}