notebooks/Postprocessing.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 348,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from pylab import *\n",
    "import seaborn as sns\n",
    "import os\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 349,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "sns.set_context('paper')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 350,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from keras.models import model_from_json"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 351,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "json_name = '../src/architecture_2016-08-06-14-35.json'\n",
    "weight_name = '../src/model_weights_2016-08-06-14-35.h5'\n",
    "model = model_from_json(open(json_name).read())\n",
    "model.load_weights(weight_name)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 352,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Model structure"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 353,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from IPython.display import SVG\n",
    "from keras.utils.visualize_util import model_to_dot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 354,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/svg+xml": [
       "<svg height=\"702pt\" viewBox=\"0.00 0.00 226.00 702.00\" width=\"226pt\" xmlns=\"http://www.w3.org/2000/svg\" xmlns:xlink=\"http://www.w3.org/1999/xlink\">\n",
       "<g class=\"graph\" id=\"graph0\" transform=\"scale(1 1) rotate(0) translate(4 698)\">\n",
       "<title>G</title>\n",
       "<polygon fill=\"white\" points=\"-4,4 -4,-698 222,-698 222,4 -4,4\" stroke=\"none\"/>\n",
       "<!-- 140676067278608 -->\n",
       "<g class=\"node\" id=\"node1\"><title>140676067278608</title>\n",
       "<polygon fill=\"none\" points=\"0,-657.5 0,-693.5 218,-693.5 218,-657.5 0,-657.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"109\" y=\"-671.8\">zeropadding2d_input_5 (InputLayer)</text>\n",
       "</g>\n",
       "<!-- 140676067278352 -->\n",
       "<g class=\"node\" id=\"node2\"><title>140676067278352</title>\n",
       "<polygon fill=\"none\" points=\"4.5,-584.5 4.5,-620.5 213.5,-620.5 213.5,-584.5 4.5,-584.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"109\" y=\"-598.8\">zeropadding2d_1 (ZeroPadding2D)</text>\n",
       "</g>\n",
       "<!-- 140676067278608&#45;&gt;140676067278352 -->\n",
       "<g class=\"edge\" id=\"edge1\"><title>140676067278608-&gt;140676067278352</title>\n",
       "<path d=\"M109,-657.313C109,-649.289 109,-639.547 109,-630.569\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"112.5,-630.529 109,-620.529 105.5,-630.529 112.5,-630.529\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140675344353552 -->\n",
       "<g class=\"node\" id=\"node3\"><title>140675344353552</title>\n",
       "<polygon fill=\"none\" points=\"6.5,-511.5 6.5,-547.5 211.5,-547.5 211.5,-511.5 6.5,-511.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"109\" y=\"-525.8\">convolution2d_1 (Convolution2D)</text>\n",
       "</g>\n",
       "<!-- 140676067278352&#45;&gt;140675344353552 -->\n",
       "<g class=\"edge\" id=\"edge2\"><title>140676067278352-&gt;140675344353552</title>\n",
       "<path d=\"M109,-584.313C109,-576.289 109,-566.547 109,-557.569\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"112.5,-557.529 109,-547.529 105.5,-557.529 112.5,-557.529\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140676068725840 -->\n",
       "<g class=\"node\" id=\"node4\"><title>140676068725840</title>\n",
       "<polygon fill=\"none\" points=\"4.5,-438.5 4.5,-474.5 213.5,-474.5 213.5,-438.5 4.5,-438.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"109\" y=\"-452.8\">zeropadding2d_2 (ZeroPadding2D)</text>\n",
       "</g>\n",
       "<!-- 140675344353552&#45;&gt;140676068725840 -->\n",
       "<g class=\"edge\" id=\"edge3\"><title>140675344353552-&gt;140676068725840</title>\n",
       "<path d=\"M109,-511.313C109,-503.289 109,-493.547 109,-484.569\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"112.5,-484.529 109,-474.529 105.5,-484.529 112.5,-484.529\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140676068745616 -->\n",
       "<g class=\"node\" id=\"node5\"><title>140676068745616</title>\n",
       "<polygon fill=\"none\" points=\"6.5,-365.5 6.5,-401.5 211.5,-401.5 211.5,-365.5 6.5,-365.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"109\" y=\"-379.8\">convolution2d_2 (Convolution2D)</text>\n",
       "</g>\n",
       "<!-- 140676068725840&#45;&gt;140676068745616 -->\n",
       "<g class=\"edge\" id=\"edge4\"><title>140676068725840-&gt;140676068745616</title>\n",
       "<path d=\"M109,-438.313C109,-430.289 109,-420.547 109,-411.569\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"112.5,-411.529 109,-401.529 105.5,-411.529 112.5,-411.529\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140677241268752 -->\n",
       "<g class=\"node\" id=\"node6\"><title>140677241268752</title>\n",
       "<polygon fill=\"none\" points=\"6,-292.5 6,-328.5 212,-328.5 212,-292.5 6,-292.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"109\" y=\"-306.8\">maxpooling2d_1 (MaxPooling2D)</text>\n",
       "</g>\n",
       "<!-- 140676068745616&#45;&gt;140677241268752 -->\n",
       "<g class=\"edge\" id=\"edge5\"><title>140676068745616-&gt;140677241268752</title>\n",
       "<path d=\"M109,-365.313C109,-357.289 109,-347.547 109,-338.569\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"112.5,-338.529 109,-328.529 105.5,-338.529 112.5,-338.529\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140676068645392 -->\n",
       "<g class=\"node\" id=\"node7\"><title>140676068645392</title>\n",
       "<polygon fill=\"none\" points=\"51.5,-219.5 51.5,-255.5 166.5,-255.5 166.5,-219.5 51.5,-219.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"109\" y=\"-233.8\">flatten_1 (Flatten)</text>\n",
       "</g>\n",
       "<!-- 140677241268752&#45;&gt;140676068645392 -->\n",
       "<g class=\"edge\" id=\"edge6\"><title>140677241268752-&gt;140676068645392</title>\n",
       "<path d=\"M109,-292.313C109,-284.289 109,-274.547 109,-265.569\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"112.5,-265.529 109,-255.529 105.5,-265.529 112.5,-265.529\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140677241151888 -->\n",
       "<g class=\"node\" id=\"node8\"><title>140677241151888</title>\n",
       "<polygon fill=\"none\" points=\"55.5,-146.5 55.5,-182.5 162.5,-182.5 162.5,-146.5 55.5,-146.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"109\" y=\"-160.8\">dense_1 (Dense)</text>\n",
       "</g>\n",
       "<!-- 140676068645392&#45;&gt;140677241151888 -->\n",
       "<g class=\"edge\" id=\"edge7\"><title>140676068645392-&gt;140677241151888</title>\n",
       "<path d=\"M109,-219.313C109,-211.289 109,-201.547 109,-192.569\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"112.5,-192.529 109,-182.529 105.5,-192.529 112.5,-192.529\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140677241214032 -->\n",
       "<g class=\"node\" id=\"node9\"><title>140677241214032</title>\n",
       "<polygon fill=\"none\" points=\"44,-73.5 44,-109.5 174,-109.5 174,-73.5 44,-73.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"109\" y=\"-87.8\">dropout_1 (Dropout)</text>\n",
       "</g>\n",
       "<!-- 140677241151888&#45;&gt;140677241214032 -->\n",
       "<g class=\"edge\" id=\"edge8\"><title>140677241151888-&gt;140677241214032</title>\n",
       "<path d=\"M109,-146.313C109,-138.289 109,-128.547 109,-119.569\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"112.5,-119.529 109,-109.529 105.5,-119.529 112.5,-119.529\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140676067822160 -->\n",
       "<g class=\"node\" id=\"node10\"><title>140676067822160</title>\n",
       "<polygon fill=\"none\" points=\"55.5,-0.5 55.5,-36.5 162.5,-36.5 162.5,-0.5 55.5,-0.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"109\" y=\"-14.8\">dense_2 (Dense)</text>\n",
       "</g>\n",
       "<!-- 140677241214032&#45;&gt;140676067822160 -->\n",
       "<g class=\"edge\" id=\"edge9\"><title>140677241214032-&gt;140676067822160</title>\n",
       "<path d=\"M109,-73.3129C109,-65.2895 109,-55.5475 109,-46.5691\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"112.5,-46.5288 109,-36.5288 105.5,-46.5289 112.5,-46.5288\" stroke=\"black\"/>\n",
       "</g>\n",
       "</g>\n",
       "</svg>"
      ],
      "text/plain": [
       "<IPython.core.display.SVG object>"
      ]
     },
     "execution_count": 354,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "SVG(model_to_dot(model).create(prog='dot', format='svg'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Training history"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 355,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "history = pd.read_csv('../src/history_2016-08-06-14-35.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 356,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZwAAAEiCAYAAADNgWQ8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7ff1b2d70710>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot(history['acc'], label='train accuracy. Higher => Better')\n",
    "plot(history['val_acc'], label='validation accuracy. Higher => Better')\n",
    "legend(loc='lower right')\n",
    "xlabel('num_epoch')\n",
    "ylabel('accuracy [$\\%$]')\n",
    "savefig('accuracy.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 357,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZkAAAEjCAYAAADg9HzdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nKGRERMQ2ChkREbGNQkZERGzz/wF3zzGU87Ri2wAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7ff1b3072e50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot(history['loss'], label='train LogLoss. Lower => Better')\n",
    "plot(history['val_loss'], label='validation LogLoss. Lower => Better')\n",
    "legend(loc='lower right')\n",
    "xlabel('num_epoch')\n",
    "ylabel('LogLoss')\n",
    "yscale('log')\n",
    "savefig('LogLoss.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "# Model performance on the test set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 358,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from keras.preprocessing.image import ImageDataGenerator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 359,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 6393 images belonging to 2 classes.\n"
     ]
    }
   ],
   "source": [
    "img_rows, img_cols = 64, 64\n",
    "\n",
    "test_datagen = ImageDataGenerator(rescale=1.0 / 255)\n",
    "\n",
    "test_path = '../data/test'\n",
    "test_generator = test_datagen.flow_from_directory(\n",
    "                                                        test_path,\n",
    "                                                        target_size=(img_rows, img_cols),\n",
    "                                                        class_mode='binary')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 360,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "(log_loss_score, accuracy_score) = model.evaluate_generator(test_generator, test_generator.N)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 361,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "('log_loss_score = ', 0.014433901161286119)"
      ]
     },
     "execution_count": 361,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "'log_loss_score = ', log_loss_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 362,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "('accuracy_score = ', 0.99405599874863126)"
      ]
     },
     "execution_count": 362,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "'accuracy_score = ', accuracy_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 363,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from sklearn.metrics import roc_auc_score, roc_curve, log_loss, accuracy_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 364,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def extract_datetime(x):\n",
    "    return x.split('.')[1].split('+')[0]    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 365,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "test_path = '../data/test'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 366,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "up_test_path = os.path.join(test_path, 'up')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 367,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "down_test_path = os.path.join(test_path, 'down')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 368,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "result = []\n",
    "up_files = os.listdir(up_test_path)\n",
    "for file_name in up_files:\n",
    "    result += [cv2.imread(os.path.join(up_test_path, file_name))]\n",
    "    \n",
    "down_files = os.listdir(down_test_path)\n",
    "for file_name in down_files:\n",
    "    result += [cv2.imread(os.path.join(down_test_path, file_name))]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 369,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "result = np.array(result).transpose((0, 3, 1, 2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 370,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "prediction = model.predict(result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 371,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df = pd.DataFrame()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 372,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df['filename'] = up_files + down_files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 373,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df['prediction'] = prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 374,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "df['state'] = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 375,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "df.loc[:len(up_files), 'state'] = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 376,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "df['datetime'] = pd.to_datetime(df['filename'].apply(extract_datetime))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 377,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "df = df.sort_values(by='datetime')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 378,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python2.7/dist-packages/ipykernel/__main__.py:1: FutureWarning: pd.rolling_mean is deprecated for ndarrays and will be removed in a future version\n",
      "  if __name__ == '__main__':\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYwAAAEnCAYAAAC64nRnAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7ff1f850b650>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot(df['datetime'].values[:2000], pd.rolling_mean(df['prediction'].values[:2000], 20), 'd-', label='prediction')\n",
    "plot(df['datetime'].values[:2000], df['state'].values[:2000], 'p-', label='state')\n",
    "xticks(rotation=45)\n",
    "legend(loc='upper right')\n",
    "savefig('rolling_mean.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 379,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.98593182957373837"
      ]
     },
     "execution_count": 379,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "roc_auc_score(df['state'].values, df['prediction'])"
   ]
  },
  {
   "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": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}