notebook/lgb_model.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "import lightgbm as lgb\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True\n"
     ]
    }
   ],
   "source": [
    "train_df = pd.read_csv('lgb_train.tsv',sep='\\t')\n",
    "test_df = pd.read_csv('lgb_test.tsv',sep='\\t')\n",
    "\n",
    "id_vars = ['guid/image']\n",
    "label_vars = ['gt_label']\n",
    "gt_vars = []\n",
    "cocov1_vars = []\n",
    "kitti_vars = []\n",
    "incep_vars = []\n",
    "for col in train_df.columns:\n",
    "    if col.startswith('gt') and col != 'gt_label':\n",
    "        gt_vars.append(col)\n",
    "    elif col.startswith('cocov1'):\n",
    "        cocov1_vars.append(col)\n",
    "    elif col.startswith('kitti'):\n",
    "        kitti_vars.append(col)\n",
    "    elif col.startswith('incep'):\n",
    "        incep_vars.append(col)\n",
    "    \n",
    "print len(gt_vars)+len(cocov1_vars)+len(kitti_vars)+len(incep_vars)+len(id_vars)+len(label_vars) ==len(train_df.columns)\n",
    "\n",
    "train_X = train_df[cocov1_vars + kitti_vars + incep_vars]\n",
    "train_y = train_df[label_vars]\n",
    "test_X = test_df[cocov1_vars + kitti_vars + incep_vars]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "((6058, 30), (1515, 30), (2631, 30))\n"
     ]
    }
   ],
   "source": [
    "# Splitting the data for model training#\n",
    "shuffle_idx = np.arange(len(train_X))\n",
    "np.random.seed(2018)\n",
    "np.random.shuffle(shuffle_idx)\n",
    "split = int(0.8*len(shuffle_idx))\n",
    "dev_X = train_X.iloc[shuffle_idx[:split],:]\n",
    "val_X = train_X.iloc[shuffle_idx[split:],:]\n",
    "dev_y = train_y.iloc[shuffle_idx[:split],:]\n",
    "val_y = train_y.iloc[shuffle_idx[split:],:]\n",
    "print(dev_X.shape, val_X.shape, test_X.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cocov1-can1-label</th>\n",
       "      <th>cocov1-can1-score</th>\n",
       "      <th>cocov1-can1-ymin</th>\n",
       "      <th>cocov1-can1-xmin</th>\n",
       "      <th>cocov1-can1-ymax</th>\n",
       "      <th>cocov1-can1-xmax</th>\n",
       "      <th>cocov1-can1-box-width</th>\n",
       "      <th>cocov1-can1-box-height</th>\n",
       "      <th>cocov1-can1-box-ratio</th>\n",
       "      <th>cocov1-can1-box-area</th>\n",
       "      <th>...</th>\n",
       "      <th>incep-can1-label</th>\n",
       "      <th>incep-can1-score</th>\n",
       "      <th>incep-can1-ymin</th>\n",
       "      <th>incep-can1-xmin</th>\n",
       "      <th>incep-can1-ymax</th>\n",
       "      <th>incep-can1-xmax</th>\n",
       "      <th>incep-can1-box-width</th>\n",
       "      <th>incep-can1-box-height</th>\n",
       "      <th>incep-can1-box-ratio</th>\n",
       "      <th>incep-can1-box-area</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>5726</th>\n",
       "      <td>2</td>\n",
       "      <td>0.999820</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>985.140326</td>\n",
       "      <td>569.375555</td>\n",
       "      <td>569.375555</td>\n",
       "      <td>985.140326</td>\n",
       "      <td>1.730212</td>\n",
       "      <td>560914.819315</td>\n",
       "      <td>...</td>\n",
       "      <td>2</td>\n",
       "      <td>0.999988</td>\n",
       "      <td>2.847549</td>\n",
       "      <td>4.005978</td>\n",
       "      <td>973.143090</td>\n",
       "      <td>536.115427</td>\n",
       "      <td>532.109449</td>\n",
       "      <td>970.295541</td>\n",
       "      <td>1.823489</td>\n",
       "      <td>516303.426171</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6931</th>\n",
       "      <td>1</td>\n",
       "      <td>0.991420</td>\n",
       "      <td>444.221048</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>560.755119</td>\n",
       "      <td>67.163571</td>\n",
       "      <td>67.163571</td>\n",
       "      <td>116.534071</td>\n",
       "      <td>1.735079</td>\n",
       "      <td>7826.844328</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0.999988</td>\n",
       "      <td>447.097097</td>\n",
       "      <td>1.281633</td>\n",
       "      <td>561.344977</td>\n",
       "      <td>72.518619</td>\n",
       "      <td>71.236986</td>\n",
       "      <td>114.247880</td>\n",
       "      <td>1.603772</td>\n",
       "      <td>8138.674625</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>1</td>\n",
       "      <td>0.999823</td>\n",
       "      <td>409.773367</td>\n",
       "      <td>1123.098311</td>\n",
       "      <td>443.245373</td>\n",
       "      <td>1187.612412</td>\n",
       "      <td>64.514101</td>\n",
       "      <td>33.472006</td>\n",
       "      <td>0.518832</td>\n",
       "      <td>2159.416335</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0.999996</td>\n",
       "      <td>407.646006</td>\n",
       "      <td>1138.056114</td>\n",
       "      <td>439.936572</td>\n",
       "      <td>1190.475788</td>\n",
       "      <td>52.419675</td>\n",
       "      <td>32.290567</td>\n",
       "      <td>0.616001</td>\n",
       "      <td>1692.660993</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3516</th>\n",
       "      <td>1</td>\n",
       "      <td>0.999376</td>\n",
       "      <td>385.345674</td>\n",
       "      <td>912.463994</td>\n",
       "      <td>429.936870</td>\n",
       "      <td>985.110351</td>\n",
       "      <td>72.646357</td>\n",
       "      <td>44.591196</td>\n",
       "      <td>0.613812</td>\n",
       "      <td>3239.387920</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0.999849</td>\n",
       "      <td>394.100230</td>\n",
       "      <td>905.887663</td>\n",
       "      <td>429.250260</td>\n",
       "      <td>984.619108</td>\n",
       "      <td>78.731445</td>\n",
       "      <td>35.150030</td>\n",
       "      <td>0.446455</td>\n",
       "      <td>2767.412640</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5040</th>\n",
       "      <td>1</td>\n",
       "      <td>0.999587</td>\n",
       "      <td>400.023007</td>\n",
       "      <td>937.893502</td>\n",
       "      <td>461.179745</td>\n",
       "      <td>1008.920218</td>\n",
       "      <td>71.026716</td>\n",
       "      <td>61.156738</td>\n",
       "      <td>0.861039</td>\n",
       "      <td>4343.762248</td>\n",
       "      <td>...</td>\n",
       "      <td>1</td>\n",
       "      <td>0.999995</td>\n",
       "      <td>407.601580</td>\n",
       "      <td>950.829748</td>\n",
       "      <td>461.936709</td>\n",
       "      <td>1016.016427</td>\n",
       "      <td>65.186679</td>\n",
       "      <td>54.335129</td>\n",
       "      <td>0.833531</td>\n",
       "      <td>3541.926568</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 30 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      cocov1-can1-label  cocov1-can1-score  cocov1-can1-ymin  \\\n",
       "5726                  2           0.999820          0.000000   \n",
       "6931                  1           0.991420        444.221048   \n",
       "16                    1           0.999823        409.773367   \n",
       "3516                  1           0.999376        385.345674   \n",
       "5040                  1           0.999587        400.023007   \n",
       "\n",
       "      cocov1-can1-xmin  cocov1-can1-ymax  cocov1-can1-xmax  \\\n",
       "5726          0.000000        985.140326        569.375555   \n",
       "6931          0.000000        560.755119         67.163571   \n",
       "16         1123.098311        443.245373       1187.612412   \n",
       "3516        912.463994        429.936870        985.110351   \n",
       "5040        937.893502        461.179745       1008.920218   \n",
       "\n",
       "      cocov1-can1-box-width  cocov1-can1-box-height  cocov1-can1-box-ratio  \\\n",
       "5726             569.375555              985.140326               1.730212   \n",
       "6931              67.163571              116.534071               1.735079   \n",
       "16                64.514101               33.472006               0.518832   \n",
       "3516              72.646357               44.591196               0.613812   \n",
       "5040              71.026716               61.156738               0.861039   \n",
       "\n",
       "      cocov1-can1-box-area         ...           incep-can1-label  \\\n",
       "5726         560914.819315         ...                          2   \n",
       "6931           7826.844328         ...                          1   \n",
       "16             2159.416335         ...                          1   \n",
       "3516           3239.387920         ...                          1   \n",
       "5040           4343.762248         ...                          1   \n",
       "\n",
       "      incep-can1-score  incep-can1-ymin  incep-can1-xmin  incep-can1-ymax  \\\n",
       "5726          0.999988         2.847549         4.005978       973.143090   \n",
       "6931          0.999988       447.097097         1.281633       561.344977   \n",
       "16            0.999996       407.646006      1138.056114       439.936572   \n",
       "3516          0.999849       394.100230       905.887663       429.250260   \n",
       "5040          0.999995       407.601580       950.829748       461.936709   \n",
       "\n",
       "      incep-can1-xmax  incep-can1-box-width  incep-can1-box-height  \\\n",
       "5726       536.115427            532.109449             970.295541   \n",
       "6931        72.518619             71.236986             114.247880   \n",
       "16        1190.475788             52.419675              32.290567   \n",
       "3516       984.619108             78.731445              35.150030   \n",
       "5040      1016.016427             65.186679              54.335129   \n",
       "\n",
       "      incep-can1-box-ratio  incep-can1-box-area  \n",
       "5726              1.823489        516303.426171  \n",
       "6931              1.603772          8138.674625  \n",
       "16                0.616001          1692.660993  \n",
       "3516              0.446455          2767.412640  \n",
       "5040              0.833531          3541.926568  \n",
       "\n",
       "[5 rows x 30 columns]"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dev_X.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1]\tvalid_0's multi_logloss: 0.956177\n",
      "Training until validation scores don't improve for 100 rounds.\n",
      "[2]\tvalid_0's multi_logloss: 0.838254\n",
      "[3]\tvalid_0's multi_logloss: 0.738915\n",
      "[4]\tvalid_0's multi_logloss: 0.654276\n",
      "[5]\tvalid_0's multi_logloss: 0.581364\n",
      "[6]\tvalid_0's multi_logloss: 0.518227\n",
      "[7]\tvalid_0's multi_logloss: 0.463006\n",
      "[8]\tvalid_0's multi_logloss: 0.414779\n",
      "[9]\tvalid_0's multi_logloss: 0.37224\n",
      "[10]\tvalid_0's multi_logloss: 0.334672\n",
      "[11]\tvalid_0's multi_logloss: 0.301412\n",
      "[12]\tvalid_0's multi_logloss: 0.271857\n",
      "[13]\tvalid_0's multi_logloss: 0.24554\n",
      "[14]\tvalid_0's multi_logloss: 0.222011\n",
      "[15]\tvalid_0's multi_logloss: 0.201041\n",
      "[16]\tvalid_0's multi_logloss: 0.182282\n",
      "[17]\tvalid_0's multi_logloss: 0.16539\n",
      "[18]\tvalid_0's multi_logloss: 0.150322\n",
      "[19]\tvalid_0's multi_logloss: 0.136756\n",
      "[20]\tvalid_0's multi_logloss: 0.124658\n",
      "[21]\tvalid_0's multi_logloss: 0.113721\n",
      "[22]\tvalid_0's multi_logloss: 0.103899\n",
      "[23]\tvalid_0's multi_logloss: 0.0951266\n",
      "[24]\tvalid_0's multi_logloss: 0.0871255\n",
      "[25]\tvalid_0's multi_logloss: 0.0800081\n",
      "[26]\tvalid_0's multi_logloss: 0.0735215\n",
      "[27]\tvalid_0's multi_logloss: 0.0675759\n",
      "[28]\tvalid_0's multi_logloss: 0.0624109\n",
      "[29]\tvalid_0's multi_logloss: 0.0577638\n",
      "[30]\tvalid_0's multi_logloss: 0.0534633\n",
      "[31]\tvalid_0's multi_logloss: 0.0497347\n",
      "[32]\tvalid_0's multi_logloss: 0.0462082\n",
      "[33]\tvalid_0's multi_logloss: 0.0429435\n",
      "[34]\tvalid_0's multi_logloss: 0.040168\n",
      "[35]\tvalid_0's multi_logloss: 0.0376284\n",
      "[36]\tvalid_0's multi_logloss: 0.0353189\n",
      "[37]\tvalid_0's multi_logloss: 0.0331974\n",
      "[38]\tvalid_0's multi_logloss: 0.0313647\n",
      "[39]\tvalid_0's multi_logloss: 0.0296295\n",
      "[40]\tvalid_0's multi_logloss: 0.0281407\n",
      "[41]\tvalid_0's multi_logloss: 0.0267925\n",
      "[42]\tvalid_0's multi_logloss: 0.0255973\n",
      "[43]\tvalid_0's multi_logloss: 0.0244412\n",
      "[44]\tvalid_0's multi_logloss: 0.023391\n",
      "[45]\tvalid_0's multi_logloss: 0.0225016\n",
      "[46]\tvalid_0's multi_logloss: 0.0216587\n",
      "[47]\tvalid_0's multi_logloss: 0.0208922\n",
      "[48]\tvalid_0's multi_logloss: 0.0202767\n",
      "[49]\tvalid_0's multi_logloss: 0.0197489\n",
      "[50]\tvalid_0's multi_logloss: 0.0192322\n",
      "[51]\tvalid_0's multi_logloss: 0.0187307\n",
      "[52]\tvalid_0's multi_logloss: 0.0182548\n",
      "[53]\tvalid_0's multi_logloss: 0.0178904\n",
      "[54]\tvalid_0's multi_logloss: 0.0175769\n",
      "[55]\tvalid_0's multi_logloss: 0.0173179\n",
      "[56]\tvalid_0's multi_logloss: 0.0171232\n",
      "[57]\tvalid_0's multi_logloss: 0.0169505\n",
      "[58]\tvalid_0's multi_logloss: 0.0168213\n",
      "[59]\tvalid_0's multi_logloss: 0.0166917\n",
      "[60]\tvalid_0's multi_logloss: 0.0165764\n",
      "[61]\tvalid_0's multi_logloss: 0.016548\n",
      "[62]\tvalid_0's multi_logloss: 0.0164842\n",
      "[63]\tvalid_0's multi_logloss: 0.0164595\n",
      "[64]\tvalid_0's multi_logloss: 0.0164712\n",
      "[65]\tvalid_0's multi_logloss: 0.0164592\n",
      "[66]\tvalid_0's multi_logloss: 0.0164389\n",
      "[67]\tvalid_0's multi_logloss: 0.0163474\n",
      "[68]\tvalid_0's multi_logloss: 0.0163266\n",
      "[69]\tvalid_0's multi_logloss: 0.0162829\n",
      "[70]\tvalid_0's multi_logloss: 0.0163015\n",
      "[71]\tvalid_0's multi_logloss: 0.0163389\n",
      "[72]\tvalid_0's multi_logloss: 0.0164373\n",
      "[73]\tvalid_0's multi_logloss: 0.0164671\n",
      "[74]\tvalid_0's multi_logloss: 0.0164842\n",
      "[75]\tvalid_0's multi_logloss: 0.0165904\n",
      "[76]\tvalid_0's multi_logloss: 0.0167127\n",
      "[77]\tvalid_0's multi_logloss: 0.0168321\n",
      "[78]\tvalid_0's multi_logloss: 0.0168426\n",
      "[79]\tvalid_0's multi_logloss: 0.0170024\n",
      "[80]\tvalid_0's multi_logloss: 0.0171435\n",
      "[81]\tvalid_0's multi_logloss: 0.0173006\n",
      "[82]\tvalid_0's multi_logloss: 0.0173898\n",
      "[83]\tvalid_0's multi_logloss: 0.0174402\n",
      "[84]\tvalid_0's multi_logloss: 0.0176575\n",
      "[85]\tvalid_0's multi_logloss: 0.0176516\n",
      "[86]\tvalid_0's multi_logloss: 0.0177816\n",
      "[87]\tvalid_0's multi_logloss: 0.0179958\n",
      "[88]\tvalid_0's multi_logloss: 0.0180323\n",
      "[89]\tvalid_0's multi_logloss: 0.0180598\n",
      "[90]\tvalid_0's multi_logloss: 0.018158\n",
      "[91]\tvalid_0's multi_logloss: 0.0181215\n",
      "[92]\tvalid_0's multi_logloss: 0.0180969\n",
      "[93]\tvalid_0's multi_logloss: 0.0181966\n",
      "[94]\tvalid_0's multi_logloss: 0.0181752\n",
      "[95]\tvalid_0's multi_logloss: 0.0181951\n",
      "[96]\tvalid_0's multi_logloss: 0.0182349\n",
      "[97]\tvalid_0's multi_logloss: 0.0182734\n",
      "[98]\tvalid_0's multi_logloss: 0.0182904\n",
      "[99]\tvalid_0's multi_logloss: 0.0182831\n",
      "[100]\tvalid_0's multi_logloss: 0.0183417\n",
      "[101]\tvalid_0's multi_logloss: 0.0183106\n",
      "[102]\tvalid_0's multi_logloss: 0.018334\n",
      "[103]\tvalid_0's multi_logloss: 0.0183636\n",
      "[104]\tvalid_0's multi_logloss: 0.0184662\n",
      "[105]\tvalid_0's multi_logloss: 0.0185748\n",
      "[106]\tvalid_0's multi_logloss: 0.0186626\n",
      "[107]\tvalid_0's multi_logloss: 0.0188351\n",
      "[108]\tvalid_0's multi_logloss: 0.0189031\n",
      "[109]\tvalid_0's multi_logloss: 0.0189397\n",
      "[110]\tvalid_0's multi_logloss: 0.019079\n",
      "[111]\tvalid_0's multi_logloss: 0.0191841\n",
      "[112]\tvalid_0's multi_logloss: 0.0192885\n",
      "[113]\tvalid_0's multi_logloss: 0.019464\n",
      "[114]\tvalid_0's multi_logloss: 0.0196504\n",
      "[115]\tvalid_0's multi_logloss: 0.0198249\n",
      "[116]\tvalid_0's multi_logloss: 0.0199213\n",
      "[117]\tvalid_0's multi_logloss: 0.0200917\n",
      "[118]\tvalid_0's multi_logloss: 0.0201823\n",
      "[119]\tvalid_0's multi_logloss: 0.0203159\n",
      "[120]\tvalid_0's multi_logloss: 0.0204497\n",
      "[121]\tvalid_0's multi_logloss: 0.0205451\n",
      "[122]\tvalid_0's multi_logloss: 0.020757\n",
      "[123]\tvalid_0's multi_logloss: 0.0208662\n",
      "[124]\tvalid_0's multi_logloss: 0.0208898\n",
      "[125]\tvalid_0's multi_logloss: 0.0209956\n",
      "[126]\tvalid_0's multi_logloss: 0.0209927\n",
      "[127]\tvalid_0's multi_logloss: 0.0211115\n",
      "[128]\tvalid_0's multi_logloss: 0.0212284\n",
      "[129]\tvalid_0's multi_logloss: 0.021346\n",
      "[130]\tvalid_0's multi_logloss: 0.0214949\n",
      "[131]\tvalid_0's multi_logloss: 0.0216809\n",
      "[132]\tvalid_0's multi_logloss: 0.0218225\n",
      "[133]\tvalid_0's multi_logloss: 0.0220246\n",
      "[134]\tvalid_0's multi_logloss: 0.0221001\n",
      "[135]\tvalid_0's multi_logloss: 0.0221959\n",
      "[136]\tvalid_0's multi_logloss: 0.0222236\n",
      "[137]\tvalid_0's multi_logloss: 0.0222462\n",
      "[138]\tvalid_0's multi_logloss: 0.0224359\n",
      "[139]\tvalid_0's multi_logloss: 0.0225228\n",
      "[140]\tvalid_0's multi_logloss: 0.022688\n",
      "[141]\tvalid_0's multi_logloss: 0.0228317\n",
      "[142]\tvalid_0's multi_logloss: 0.0229498\n",
      "[143]\tvalid_0's multi_logloss: 0.0230609\n",
      "[144]\tvalid_0's multi_logloss: 0.0230926\n",
      "[145]\tvalid_0's multi_logloss: 0.0232594\n",
      "[146]\tvalid_0's multi_logloss: 0.0233223\n",
      "[147]\tvalid_0's multi_logloss: 0.0234361\n",
      "[148]\tvalid_0's multi_logloss: 0.0236497\n",
      "[149]\tvalid_0's multi_logloss: 0.0237611\n",
      "[150]\tvalid_0's multi_logloss: 0.0238891\n",
      "[151]\tvalid_0's multi_logloss: 0.0240424\n",
      "[152]\tvalid_0's multi_logloss: 0.0242145\n",
      "[153]\tvalid_0's multi_logloss: 0.0243228\n",
      "[154]\tvalid_0's multi_logloss: 0.0244952\n",
      "[155]\tvalid_0's multi_logloss: 0.0246081\n",
      "[156]\tvalid_0's multi_logloss: 0.0247179\n",
      "[157]\tvalid_0's multi_logloss: 0.0248484\n",
      "[158]\tvalid_0's multi_logloss: 0.0250191\n",
      "[159]\tvalid_0's multi_logloss: 0.0250286\n",
      "[160]\tvalid_0's multi_logloss: 0.0251494\n",
      "[161]\tvalid_0's multi_logloss: 0.0251545\n",
      "[162]\tvalid_0's multi_logloss: 0.0252647\n",
      "[163]\tvalid_0's multi_logloss: 0.0253663\n",
      "[164]\tvalid_0's multi_logloss: 0.0253763\n",
      "[165]\tvalid_0's multi_logloss: 0.025594\n",
      "[166]\tvalid_0's multi_logloss: 0.0256951\n",
      "[167]\tvalid_0's multi_logloss: 0.0257841\n",
      "[168]\tvalid_0's multi_logloss: 0.0258163\n",
      "[169]\tvalid_0's multi_logloss: 0.0258701\n",
      "Early stopping, best iteration is:\n",
      "[69]\tvalid_0's multi_logloss: 0.0162829\n"
     ]
    }
   ],
   "source": [
    "def run_lgb(train_X, train_y, val_X, val_y, test_X):\n",
    "\n",
    "    lgtrain = lgb.Dataset(train_X, label=train_y, categorical_feature=['cocov1-can1-label','kitti-can1-label'])\n",
    "    lgval = lgb.Dataset(val_X, label=val_y, reference=lgtrain, categorical_feature=['cocov1-can1-label','kitti-can1-label'])\n",
    "\n",
    "    evals_result = {}\n",
    "    params = {\n",
    "        'objective' : 'multiclass',\n",
    "        'num_class' : 3,\n",
    "        'learning_rate' : 0.1,\n",
    "        'bagging_seed' : 2018,\n",
    "    }\n",
    "    model = lgb.train(params, \n",
    "                      lgtrain, \n",
    "                      num_boost_round=1000, \n",
    "                      valid_sets=[lgval], \n",
    "                      categorical_feature=['cocov1-can1-label','kitti-can1-label'],\n",
    "                      early_stopping_rounds=100,\n",
    "                      evals_result=evals_result)\n",
    "    \n",
    "    pred_test_y = model.predict(test_X, num_iteration=model.best_iteration)\n",
    "    return pred_test_y, model, evals_result\n",
    "    \n",
    "    \n",
    "pred_test, model, evals_result = run_lgb(dev_X, dev_y, val_X, val_y, test_X)\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11568d450>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(12,18))\n",
    "lgb.plot_importance(model, height=0.8, ax=ax)\n",
    "ax.grid(False)\n",
    "plt.title(\"LightGBM - Feature Importance\", fontsize=15)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cocov1-can1-label</th>\n",
       "      <th>cocov1-can1-score</th>\n",
       "      <th>cocov1-can1-ymin</th>\n",
       "      <th>cocov1-can1-xmin</th>\n",
       "      <th>cocov1-can1-ymax</th>\n",
       "      <th>cocov1-can1-xmax</th>\n",
       "      <th>cocov1-can1-box-width</th>\n",
       "      <th>cocov1-can1-box-height</th>\n",
       "      <th>cocov1-can1-box-ratio</th>\n",
       "      <th>cocov1-can1-box-area</th>\n",
       "      <th>...</th>\n",
       "      <th>incep-can1-label</th>\n",
       "      <th>incep-can1-score</th>\n",
       "      <th>incep-can1-ymin</th>\n",
       "      <th>incep-can1-xmin</th>\n",
       "      <th>incep-can1-ymax</th>\n",
       "      <th>incep-can1-xmax</th>\n",
       "      <th>incep-can1-box-width</th>\n",
       "      <th>incep-can1-box-height</th>\n",
       "      <th>incep-can1-box-ratio</th>\n",
       "      <th>incep-can1-box-area</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>105.870206</td>\n",
       "      <td>778.004116</td>\n",
       "      <td>599.913759</td>\n",
       "      <td>1198.261631</td>\n",
       "      <td>420.257514</td>\n",
       "      <td>494.043552</td>\n",
       "      <td>1.175573</td>\n",
       "      <td>207625.515156</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0.999922</td>\n",
       "      <td>114.727974</td>\n",
       "      <td>778.541677</td>\n",
       "      <td>597.486985</td>\n",
       "      <td>1199.429844</td>\n",
       "      <td>420.888167</td>\n",
       "      <td>482.759011</td>\n",
       "      <td>1.147001</td>\n",
       "      <td>203187.555194</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>107.833644</td>\n",
       "      <td>768.235507</td>\n",
       "      <td>598.237929</td>\n",
       "      <td>1182.570957</td>\n",
       "      <td>414.335451</td>\n",
       "      <td>490.404285</td>\n",
       "      <td>1.183592</td>\n",
       "      <td>203191.880580</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0.999919</td>\n",
       "      <td>118.921772</td>\n",
       "      <td>768.581806</td>\n",
       "      <td>588.000484</td>\n",
       "      <td>1186.015816</td>\n",
       "      <td>417.434010</td>\n",
       "      <td>469.078712</td>\n",
       "      <td>1.123719</td>\n",
       "      <td>195809.407618</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>99.043518</td>\n",
       "      <td>796.189221</td>\n",
       "      <td>601.950952</td>\n",
       "      <td>1224.586350</td>\n",
       "      <td>428.397129</td>\n",
       "      <td>502.907433</td>\n",
       "      <td>1.173928</td>\n",
       "      <td>215444.100425</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0.999927</td>\n",
       "      <td>108.871812</td>\n",
       "      <td>798.997039</td>\n",
       "      <td>594.052243</td>\n",
       "      <td>1221.967910</td>\n",
       "      <td>422.970871</td>\n",
       "      <td>485.180431</td>\n",
       "      <td>1.147078</td>\n",
       "      <td>205217.189787</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>98.499827</td>\n",
       "      <td>792.350204</td>\n",
       "      <td>604.166475</td>\n",
       "      <td>1221.011094</td>\n",
       "      <td>428.660889</td>\n",
       "      <td>505.666648</td>\n",
       "      <td>1.179643</td>\n",
       "      <td>216759.515104</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0.999941</td>\n",
       "      <td>112.773276</td>\n",
       "      <td>792.317748</td>\n",
       "      <td>592.349200</td>\n",
       "      <td>1215.951043</td>\n",
       "      <td>423.633296</td>\n",
       "      <td>479.575925</td>\n",
       "      <td>1.132054</td>\n",
       "      <td>203164.329518</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>99.756652</td>\n",
       "      <td>789.944644</td>\n",
       "      <td>605.358793</td>\n",
       "      <td>1219.450434</td>\n",
       "      <td>429.505790</td>\n",
       "      <td>505.602141</td>\n",
       "      <td>1.177172</td>\n",
       "      <td>217159.047226</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0.999928</td>\n",
       "      <td>111.382233</td>\n",
       "      <td>791.457046</td>\n",
       "      <td>593.168555</td>\n",
       "      <td>1211.594545</td>\n",
       "      <td>420.137499</td>\n",
       "      <td>481.786322</td>\n",
       "      <td>1.146735</td>\n",
       "      <td>202416.500180</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 30 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   cocov1-can1-label  cocov1-can1-score  cocov1-can1-ymin  cocov1-can1-xmin  \\\n",
       "0                  0                1.0        105.870206        778.004116   \n",
       "1                  0                1.0        107.833644        768.235507   \n",
       "2                  0                1.0         99.043518        796.189221   \n",
       "3                  0                1.0         98.499827        792.350204   \n",
       "4                  0                1.0         99.756652        789.944644   \n",
       "\n",
       "   cocov1-can1-ymax  cocov1-can1-xmax  cocov1-can1-box-width  \\\n",
       "0        599.913759       1198.261631             420.257514   \n",
       "1        598.237929       1182.570957             414.335451   \n",
       "2        601.950952       1224.586350             428.397129   \n",
       "3        604.166475       1221.011094             428.660889   \n",
       "4        605.358793       1219.450434             429.505790   \n",
       "\n",
       "   cocov1-can1-box-height  cocov1-can1-box-ratio  cocov1-can1-box-area  \\\n",
       "0              494.043552               1.175573         207625.515156   \n",
       "1              490.404285               1.183592         203191.880580   \n",
       "2              502.907433               1.173928         215444.100425   \n",
       "3              505.666648               1.179643         216759.515104   \n",
       "4              505.602141               1.177172         217159.047226   \n",
       "\n",
       "          ...           incep-can1-label  incep-can1-score  incep-can1-ymin  \\\n",
       "0         ...                          0          0.999922       114.727974   \n",
       "1         ...                          0          0.999919       118.921772   \n",
       "2         ...                          0          0.999927       108.871812   \n",
       "3         ...                          0          0.999941       112.773276   \n",
       "4         ...                          0          0.999928       111.382233   \n",
       "\n",
       "   incep-can1-xmin  incep-can1-ymax  incep-can1-xmax  incep-can1-box-width  \\\n",
       "0       778.541677       597.486985      1199.429844            420.888167   \n",
       "1       768.581806       588.000484      1186.015816            417.434010   \n",
       "2       798.997039       594.052243      1221.967910            422.970871   \n",
       "3       792.317748       592.349200      1215.951043            423.633296   \n",
       "4       791.457046       593.168555      1211.594545            420.137499   \n",
       "\n",
       "   incep-can1-box-height  incep-can1-box-ratio  incep-can1-box-area  \n",
       "0             482.759011              1.147001        203187.555194  \n",
       "1             469.078712              1.123719        195809.407618  \n",
       "2             485.180431              1.147078        205217.189787  \n",
       "3             479.575925              1.132054        203164.329518  \n",
       "4             481.786322              1.146735        202416.500180  \n",
       "\n",
       "[5 rows x 30 columns]"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_X.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_df['lgb_pred'] = np.argmax(pred_test,axis=1)\n",
    "compare_df = test_df[['lgb_pred','cocov1-can1-label','kitti-can1-label']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(308, 3)"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "compare_df[compare_df['lgb_pred'] != compare_df['cocov1-can1-label']].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>guid/image</th>\n",
       "      <th>lgb_pred</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0410e7f5-4bfd-4d35-b623-9f8d6b87c179/0000</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0410e7f5-4bfd-4d35-b623-9f8d6b87c179/0001</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0410e7f5-4bfd-4d35-b623-9f8d6b87c179/0002</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0410e7f5-4bfd-4d35-b623-9f8d6b87c179/0003</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0410e7f5-4bfd-4d35-b623-9f8d6b87c179/0004</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                  guid/image  lgb_pred\n",
       "0  0410e7f5-4bfd-4d35-b623-9f8d6b87c179/0000         0\n",
       "1  0410e7f5-4bfd-4d35-b623-9f8d6b87c179/0001         0\n",
       "2  0410e7f5-4bfd-4d35-b623-9f8d6b87c179/0002         0\n",
       "3  0410e7f5-4bfd-4d35-b623-9f8d6b87c179/0003         0\n",
       "4  0410e7f5-4bfd-4d35-b623-9f8d6b87c179/0004         0"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sub = test_df[['guid/image','lgb_pred']]\n",
    "sub.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "sub.columns = ['guid/image','label']\n",
    "sub.to_csv('sub14.csv',index=False)"
   ]
  },
  {
   "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": "2.7.14"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}