readAndSubmit_sample.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from sklearn.model_selection import train_test_split,KFold\n",
    "from sklearn.ensemble import ExtraTreesClassifier,RandomForestClassifier,GradientBoostingClassifier\n",
    "from sklearn.metrics import classification_report, log_loss\n",
    "from sklearn.metrics import confusion_matrix\n",
    "\n",
    "import random\n",
    "\n",
    "random.seed(12345)\n",
    "\n",
    "pd.options.display.max_columns = 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "path = './'\n",
    "name = ''\n",
    "out_name = path + name + 'submission_py.csv'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "\n",
    "import itertools\n",
    "\n",
    "def plot_confusion_matrix(cm, classes,\n",
    "                          normalize=False,\n",
    "                          title='Confusion matrix',\n",
    "                          doprint=True,\n",
    "                          cmap=plt.cm.Blues):\n",
    "    \"\"\"\n",
    "    This function prints and plots the confusion matrix.\n",
    "    Normalization can be applied by setting `normalize=True`.\n",
    "    \"\"\"\n",
    "    if normalize:\n",
    "        cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n",
    "        print(\"Normalized confusion matrix\")\n",
    "    else:\n",
    "        print('Confusion matrix, without normalization')\n",
    "\n",
    "    if doprint:\n",
    "        print(cm)\n",
    "\n",
    "    plt.imshow(cm, interpolation='nearest', cmap=cmap)\n",
    "    plt.title(title)\n",
    "    plt.colorbar()\n",
    "    tick_marks = np.arange(len(classes))\n",
    "    plt.xticks(tick_marks, classes, rotation=45)\n",
    "    plt.yticks(tick_marks, classes)\n",
    "\n",
    "    fmt = '.2f' if normalize else 'd'\n",
    "    thresh = cm.max() / 2.\n",
    "    if doprint:\n",
    "        for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n",
    "            plt.text(j, i, format(cm[i, j], fmt),\n",
    "                     horizontalalignment=\"center\",\n",
    "                     color=\"white\" if cm[i, j] > thresh else \"black\")\n",
    "\n",
    "    plt.tight_layout()\n",
    "    plt.ylabel('True label')\n",
    "    plt.xlabel('Predicted label')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [],
   "source": [
    "# read data\n",
    "\n",
    "train_data = pd.read_csv(path + name + 'train.csv')\n",
    "test_data = pd.read_csv(path + name + 'test.csv')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(28746, 213)\n",
      "(271251, 211)\n"
     ]
    }
   ],
   "source": [
    "print(train_data.shape)\n",
    "print(test_data.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(28746, 70)\n",
      "(271251, 70)\n",
      "(28746,)\n"
     ]
    }
   ],
   "source": [
    "# get first 5 seconds\n",
    "qq = list(range(1,71))\n",
    "X = train_data.iloc[:,qq].as_matrix()\n",
    "Xt = test_data.iloc[:,qq].as_matrix()\n",
    "y = train_data.iloc[:,-1].as_matrix()\n",
    "\n",
    "print(X.shape)\n",
    "print(Xt.shape)\n",
    "print(y.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Test XGBoost on first 5 seconds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(27308, 70)\n",
      "(1438, 70)\n"
     ]
    }
   ],
   "source": [
    "X_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.05)\n",
    "print(X_train.shape)\n",
    "print(X_val.shape)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "          1       0.89      0.79      0.84        91\n",
      "          2       0.78      0.89      0.83        74\n",
      "          3       0.31      0.59      0.41        82\n",
      "          4       0.43      0.39      0.41        59\n",
      "          5       0.38      0.30      0.33        88\n",
      "          6       0.32      0.17      0.22        75\n",
      "          7       0.23      0.40      0.29        73\n",
      "          8       0.34      0.41      0.37        88\n",
      "          9       0.23      0.14      0.18        77\n",
      "         10       0.32      0.28      0.30        54\n",
      "         11       0.41      0.15      0.22        46\n",
      "         12       0.30      0.16      0.21        43\n",
      "         13       0.42      0.70      0.52        61\n",
      "         14       0.29      0.44      0.35        54\n",
      "         15       0.23      0.16      0.19        50\n",
      "         16       0.11      0.05      0.07        38\n",
      "         17       0.33      0.33      0.33        42\n",
      "         18       0.16      0.13      0.14        38\n",
      "         19       0.24      0.16      0.19        38\n",
      "         20       0.30      0.31      0.30        45\n",
      "         21       0.25      0.21      0.23        39\n",
      "         22       0.18      0.24      0.20        38\n",
      "         23       0.37      0.21      0.27        47\n",
      "         24       0.22      0.26      0.24        58\n",
      "         25       0.38      0.28      0.32        40\n",
      "\n",
      "avg / total       0.36      0.36      0.35      1438\n",
      "\n",
      "log_loss: 1.69511740974\n",
      "CPU times: user 1min 59s, sys: 28 ms, total: 2min\n",
      "Wall time: 30.2 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "from xgboost import XGBClassifier\n",
    "gbc = XGBClassifier(objective='multi:softprob',\n",
    "                    learning_rate=0.2,\n",
    "                    subsample=0.7,\n",
    "                    colsample_bytree=0.9,\n",
    "                    colsample_bylevel=0.7,\n",
    "                    max_depth=5,\n",
    "                    nthread=4,\n",
    "                    n_estimators=50,\n",
    "                    seed=1234)\n",
    "\n",
    "gbc.fit(X_train,y_train)\n",
    "pred = gbc.predict(X_val)\n",
    "print(classification_report(y_pred=pred,y_true=y_val))\n",
    "pred_proba = gbc.predict_proba(X_val)\n",
    "print('log_loss: {}'.format(log_loss(y_pred=pred_proba,y_true=y_val)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Normalized confusion matrix\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAckAAAGbCAYAAACveWuNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f97d56a2850>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "cnf_matrix = confusion_matrix(y_val, pred)\n",
    "plt.figure()\n",
    "classes = range(0,25)\n",
    "plot_confusion_matrix(cnf_matrix, classes, normalize=True, doprint=False, title='Normalized confusion matrix')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "#submit result\n",
    "\n",
    "pred = gbc.predict(Xt)\n",
    "pred = pred.astype(int)\n",
    "df = pd.DataFrame(pred)\n",
    "df.to_csv(out_name, header=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "#compress result file\n",
    "import gzip\n",
    "\n",
    "in_data = open(out_name, \"rb\").read()\n",
    "out_gz = out_name+\".gz\"\n",
    "gzf = gzip.open(out_gz, \"wb\")\n",
    "gzf.write(in_data)\n",
    "gzf.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "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": 1
}