eyn-pre-cluster-calc.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "_cell_guid": "b1076dfc-b9ad-4769-8c92-a6c4dae69d19",
    "_uuid": "8f2839f25d086af736a60e9eeb907d3b93b6e0e5"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['train_data.npy', 'trn_index_list.npy', 'custom.css', 'val_index_list.npy', '__results__.html', 'train_targets_inside.npy', 'submission.csv', 'train_targets.npy', 'test_ids.npy', '__output__.json', 'test_data.npy', '__notebook__.ipynb']\n"
     ]
    }
   ],
   "source": [
    "# place all the points on the map of last-seen-nonstat, excluding itself\n",
    "# folds information actually not necessary\n",
    "import os\n",
    "from tqdm import tqdm\n",
    "import numpy as np # linear algebra\n",
    "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n",
    "import matplotlib.pyplot as plt\n",
    "print(os.listdir(\"../input/eyn-original\"))\n",
    "x_min, x_max = -1., 1.\n",
    "y_min, y_max = -.3, .3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2815323, 40) (703815, 40)\n"
     ]
    }
   ],
   "source": [
    "# to load\n",
    "df_train_cluster = pd.read_pickle(\"../input/eyn-pre-cluster-unif/df_train_cluster.pickle\")\n",
    "df_test_cluster = pd.read_pickle(\"../input/eyn-pre-cluster-unif/df_test_cluster.pickle\")\n",
    "print(df_train_cluster.shape, df_test_cluster.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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></th>\n",
       "      <th>at0</th>\n",
       "      <th>ad0</th>\n",
       "      <th>bt0</th>\n",
       "      <th>bd0</th>\n",
       "      <th>at1</th>\n",
       "      <th>ad1</th>\n",
       "      <th>bt1</th>\n",
       "      <th>bd1</th>\n",
       "      <th>at2</th>\n",
       "      <th>ad2</th>\n",
       "      <th>bt2</th>\n",
       "      <th>bd2</th>\n",
       "      <th>at3</th>\n",
       "      <th>ad3</th>\n",
       "      <th>bt3</th>\n",
       "      <th>bd3</th>\n",
       "      <th>at4</th>\n",
       "      <th>ad4</th>\n",
       "      <th>bt4</th>\n",
       "      <th>bd4</th>\n",
       "      <th>at5</th>\n",
       "      <th>ad5</th>\n",
       "      <th>bt5</th>\n",
       "      <th>bd5</th>\n",
       "      <th>at6</th>\n",
       "      <th>ad6</th>\n",
       "      <th>bt6</th>\n",
       "      <th>bd6</th>\n",
       "      <th>at7</th>\n",
       "      <th>ad7</th>\n",
       "      <th>bt7</th>\n",
       "      <th>bd7</th>\n",
       "      <th>at8</th>\n",
       "      <th>ad8</th>\n",
       "      <th>bt8</th>\n",
       "      <th>bd8</th>\n",
       "      <th>at9</th>\n",
       "      <th>ad9</th>\n",
       "      <th>bt9</th>\n",
       "      <th>bd9</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>seq</th>\n",
       "      <th>traj</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">0</th>\n",
       "      <th>0</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001438</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001978</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.002254</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.002459</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.002761</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.003468</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.004074</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.004421</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.004642</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000724</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000724</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001074</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001074</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001173</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001173</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001282</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001282</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001336</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001336</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001341</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001341</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001341</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001341</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001809</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001809</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001930</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001930</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.002164</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.002164</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000145</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.002898</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000145</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.006580</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000147</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.006816</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000147</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.006963</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000149</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.007184</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000243</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.007247</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000268</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.007351</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000333</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.007609</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000334</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.007850</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000336</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.008035</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001014</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000649</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001377</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000758</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001447</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000960</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001587</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001008</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001615</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001141</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001627</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001560</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001855</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001563</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001967</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001708</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.002183</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001980</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.002303</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.002017</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.0</td>\n",
       "      <td>0.016330</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000426</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.022404</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000426</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.027182</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000426</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.029146</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.000736</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.030046</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001119</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.030737</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001509</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.030909</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.002791</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.031490</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.003119</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.032276</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.005444</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.032572</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.005448</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          at0       ad0  bt0       bd0    ...     at9       ad9  bt9       bd9\n",
       "seq traj                                  ...                                 \n",
       "0   0     NaN       NaN  NaN       NaN    ...     0.0  0.004642  NaN       NaN\n",
       "    1     0.0  0.000724  0.0  0.000724    ...     0.0  0.002164  0.0  0.002164\n",
       "    2     0.0  0.000145  0.0  0.002898    ...     0.0  0.000336  0.0  0.008035\n",
       "    3     0.0  0.001014  0.0  0.000649    ...     0.0  0.002303  0.0  0.002017\n",
       "    4     0.0  0.016330  0.0  0.000426    ...     0.0  0.032572  0.0  0.005448\n",
       "\n",
       "[5 rows x 40 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_train_cluster.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2815323, 40) (703815, 40)\n"
     ]
    }
   ],
   "source": [
    "df_train_cluster_values = df_train_cluster.values\n",
    "df_test_cluster_values = df_test_cluster.values\n",
    "df_original_columns = df_train_cluster.columns\n",
    "print(df_train_cluster_values.shape, df_test_cluster_values.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2815323, 10) (2815323, 10) (2815323, 10) (2815323, 10)\n",
      "(703815, 10) (703815, 10) (703815, 10) (703815, 10)\n"
     ]
    }
   ],
   "source": [
    "QUERY_SIZE = df_train_cluster_values[:,0::4].shape[1]\n",
    "\n",
    "df_train_entry_target = df_train_cluster_values[:,0::4]\n",
    "df_train_exit_target = df_train_cluster_values[:,2::4]\n",
    "df_train_entry_dist = df_train_cluster_values[:,1::4]\n",
    "df_train_exit_dist = df_train_cluster_values[:,3::4]\n",
    "print(df_train_entry_target.shape, df_train_exit_target.shape, \n",
    "      df_train_entry_dist.shape, df_train_exit_dist.shape)\n",
    "\n",
    "df_test_entry_target = df_test_cluster_values[:,0::4]\n",
    "df_test_exit_target = df_test_cluster_values[:,2::4]\n",
    "df_test_entry_dist = df_test_cluster_values[:,1::4]\n",
    "df_test_exit_dist = df_test_cluster_values[:,3::4]\n",
    "print(df_test_entry_target.shape, df_test_exit_target.shape, \n",
    "      df_test_entry_dist.shape, df_test_exit_dist.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1728x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(24,4))\n",
    "for i in range(1,10):\n",
    "    plt.hist(list(df_train_entry_dist[::21,i]), density=True, bins='scott', histtype='step')\n",
    "    plt.axvline(x=np.mean(df_train_entry_dist[::21,i]))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "DECAY_CONST = 0.001\n",
    "df_train_entry_dist_w = np.exp(-np.array(df_train_entry_dist, dtype=np.float32) / 0.001)\n",
    "df_train_exit_dist_w = np.exp(-np.array(df_train_exit_dist, dtype=np.float32) / 0.001)\n",
    "df_test_entry_dist_w = np.exp(-np.array(df_test_entry_dist, dtype=np.float32) / 0.001)\n",
    "df_test_exit_dist_w = np.exp(-np.array(df_test_exit_dist, dtype=np.float32) / 0.001)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2815323, 10) (2815323, 10) (703815, 10) (703815, 10)\n"
     ]
    }
   ],
   "source": [
    "df_train_entry_w_dot = df_train_entry_target-0.5 * df_train_entry_dist_w\n",
    "df_train_exit_w_dot = df_train_exit_target-0.5 * df_train_entry_dist_w\n",
    "df_test_entry_w_dot = df_test_entry_target-0.5 * df_test_entry_dist_w\n",
    "df_test_exit_w_dot = df_test_exit_target-0.5 * df_test_entry_dist_w\n",
    "print(df_train_entry_w_dot.shape, df_train_exit_w_dot.shape, \n",
    "      df_test_entry_w_dot.shape, df_test_exit_w_dot.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "entry_col_names_weighted = [\"c{}\".format(i) for i in range(QUERY_SIZE)]\n",
    "exit_col_names_weighted = [\"d{}\".format(i) for i in range(QUERY_SIZE)]\n",
    "\n",
    "for i,(a,b) in enumerate(zip(entry_col_names_weighted, exit_col_names_weighted)):\n",
    "    df_train_cluster[a] = df_train_entry_w_dot[:,i].ravel()\n",
    "    df_train_cluster[b] = df_train_exit_w_dot[:,i].ravel()\n",
    "    df_test_cluster[a] = df_test_entry_w_dot[:,i].ravel()\n",
    "    df_test_cluster[b] = df_test_exit_w_dot[:,i].ravel()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_train_cluster = df_train_cluster.drop(columns=df_original_columns)\n",
    "df_test_cluster = df_test_cluster.drop(columns=df_original_columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_train_cluster.to_pickle(\"df_train_cluster_w.pickle\")\n",
    "df_test_cluster.to_pickle(\"df_test_cluster_w.pickle\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "%reset -sf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['train_data.npy', 'trn_index_list.npy', 'custom.css', 'val_index_list.npy', '__results__.html', 'train_targets_inside.npy', 'submission.csv', 'train_targets.npy', 'test_ids.npy', '__output__.json', 'test_data.npy', '__notebook__.ipynb']\n"
     ]
    }
   ],
   "source": [
    "# place all the points on the map of last-seen-nonstat, excluding itself\n",
    "# folds information actually not necessary\n",
    "import os\n",
    "from tqdm import tqdm\n",
    "import numpy as np # linear algebra\n",
    "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n",
    "import matplotlib.pyplot as plt\n",
    "print(os.listdir(\"../input/eyn-original\"))\n",
    "x_min, x_max = -1., 1.\n",
    "y_min, y_max = -.3, .3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2815323, 20) (703815, 20)\n"
     ]
    },
    {
     "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></th>\n",
       "      <th>c0</th>\n",
       "      <th>d0</th>\n",
       "      <th>c1</th>\n",
       "      <th>d1</th>\n",
       "      <th>c2</th>\n",
       "      <th>d2</th>\n",
       "      <th>c3</th>\n",
       "      <th>d3</th>\n",
       "      <th>c4</th>\n",
       "      <th>d4</th>\n",
       "      <th>c5</th>\n",
       "      <th>d5</th>\n",
       "      <th>c6</th>\n",
       "      <th>d6</th>\n",
       "      <th>c7</th>\n",
       "      <th>d7</th>\n",
       "      <th>c8</th>\n",
       "      <th>d8</th>\n",
       "      <th>c9</th>\n",
       "      <th>d9</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>seq</th>\n",
       "      <th>traj</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"5\" valign=\"top\">0</th>\n",
       "      <th>0</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.118651</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.0691771</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.0524919</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.0427709</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.0316059</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.0155893</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.00850572</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.00601303</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.00481764</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>-0.242493</td>\n",
       "      <td>-0.242493</td>\n",
       "      <td>-0.170878</td>\n",
       "      <td>-0.170878</td>\n",
       "      <td>-0.154685</td>\n",
       "      <td>-0.154685</td>\n",
       "      <td>-0.13878</td>\n",
       "      <td>-0.13878</td>\n",
       "      <td>-0.131465</td>\n",
       "      <td>-0.131465</td>\n",
       "      <td>-0.130822</td>\n",
       "      <td>-0.130822</td>\n",
       "      <td>-0.130812</td>\n",
       "      <td>-0.130812</td>\n",
       "      <td>-0.0819298</td>\n",
       "      <td>-0.0819298</td>\n",
       "      <td>-0.0725769</td>\n",
       "      <td>-0.0725769</td>\n",
       "      <td>-0.0574125</td>\n",
       "      <td>-0.0574125</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-0.432722</td>\n",
       "      <td>-0.432722</td>\n",
       "      <td>-0.432425</td>\n",
       "      <td>-0.432425</td>\n",
       "      <td>-0.431771</td>\n",
       "      <td>-0.431771</td>\n",
       "      <td>-0.431658</td>\n",
       "      <td>-0.431658</td>\n",
       "      <td>-0.430884</td>\n",
       "      <td>-0.430884</td>\n",
       "      <td>-0.392217</td>\n",
       "      <td>-0.392217</td>\n",
       "      <td>-0.382518</td>\n",
       "      <td>-0.382518</td>\n",
       "      <td>-0.358415</td>\n",
       "      <td>-0.358415</td>\n",
       "      <td>-0.358167</td>\n",
       "      <td>-0.358167</td>\n",
       "      <td>-0.35735</td>\n",
       "      <td>-0.35735</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>-0.181469</td>\n",
       "      <td>-0.181469</td>\n",
       "      <td>-0.126156</td>\n",
       "      <td>-0.126156</td>\n",
       "      <td>-0.117624</td>\n",
       "      <td>-0.117624</td>\n",
       "      <td>-0.102247</td>\n",
       "      <td>-0.102247</td>\n",
       "      <td>-0.0994271</td>\n",
       "      <td>-0.0994271</td>\n",
       "      <td>-0.0982649</td>\n",
       "      <td>-0.0982649</td>\n",
       "      <td>-0.0782405</td>\n",
       "      <td>-0.0782405</td>\n",
       "      <td>-0.0699517</td>\n",
       "      <td>-0.0699517</td>\n",
       "      <td>-0.0563424</td>\n",
       "      <td>-0.0563424</td>\n",
       "      <td>-0.0499834</td>\n",
       "      <td>-0.0499834</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>-4.04356e-08</td>\n",
       "      <td>-4.04356e-08</td>\n",
       "      <td>-9.31554e-11</td>\n",
       "      <td>-9.31554e-11</td>\n",
       "      <td>-7.83267e-13</td>\n",
       "      <td>-7.83267e-13</td>\n",
       "      <td>-1.09901e-13</td>\n",
       "      <td>-1.09901e-13</td>\n",
       "      <td>-4.46805e-14</td>\n",
       "      <td>-4.46805e-14</td>\n",
       "      <td>-2.23839e-14</td>\n",
       "      <td>-2.23839e-14</td>\n",
       "      <td>-1.8855e-14</td>\n",
       "      <td>-1.8855e-14</td>\n",
       "      <td>-1.05498e-14</td>\n",
       "      <td>-1.05498e-14</td>\n",
       "      <td>-4.80605e-15</td>\n",
       "      <td>-4.80605e-15</td>\n",
       "      <td>-3.57428e-15</td>\n",
       "      <td>-3.57428e-15</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                   c0           d0     ...                c9           d9\n",
       "seq traj                               ...                               \n",
       "0   0             NaN          NaN     ...       -0.00481764          NaN\n",
       "    1       -0.242493    -0.242493     ...        -0.0574125   -0.0574125\n",
       "    2       -0.432722    -0.432722     ...          -0.35735     -0.35735\n",
       "    3       -0.181469    -0.181469     ...        -0.0499834   -0.0499834\n",
       "    4    -4.04356e-08 -4.04356e-08     ...      -3.57428e-15 -3.57428e-15\n",
       "\n",
       "[5 rows x 20 columns]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# try loading\n",
    "df_train_cluster_w = pd.read_pickle(\"df_train_cluster_w.pickle\")\n",
    "df_test_cluster_w = pd.read_pickle(\"df_test_cluster_w.pickle\")\n",
    "print(df_train_cluster_w.shape, df_test_cluster_w.shape)\n",
    "df_train_cluster_w.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_train_cluster_w_values = df_train_cluster_w.values\n",
    "df_test_cluster_w_values = df_test_cluster_w.values\n",
    "df_original_columns = df_test_cluster_w.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_train_cluster_w_sum_entry = np.sum(df_train_cluster_w_values[:,0::2], axis=1)\n",
    "df_train_cluster_w_sum_exit = np.sum(df_train_cluster_w_values[:,1::2], axis=1)\n",
    "df_test_cluster_w_sum_entry = np.sum(df_test_cluster_w_values[:,0::2], axis=1)\n",
    "df_test_cluster_w_sum_exit = np.sum(df_test_cluster_w_values[:,1::2], axis=1)\n",
    "\n",
    "df_train_cluster_w_sum_entry[::21] = np.nansum(df_train_cluster_w_values[::21,0::2], axis=1)\n",
    "df_test_cluster_w_sum_entry[::21] = np.nansum(df_test_cluster_w_values[::21,0::2], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_train_cluster_w[\"sum_entry\"] = df_train_cluster_w_sum_entry\n",
    "df_train_cluster_w[\"sum_exit\"] = df_train_cluster_w_sum_exit\n",
    "df_test_cluster_w[\"sum_entry\"] = df_test_cluster_w_sum_entry\n",
    "df_test_cluster_w[\"sum_exit\"] = df_test_cluster_w_sum_exit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "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></th>\n",
       "      <th>c0</th>\n",
       "      <th>d0</th>\n",
       "      <th>c1</th>\n",
       "      <th>d1</th>\n",
       "      <th>c2</th>\n",
       "      <th>d2</th>\n",
       "      <th>c3</th>\n",
       "      <th>d3</th>\n",
       "      <th>c4</th>\n",
       "      <th>d4</th>\n",
       "      <th>c5</th>\n",
       "      <th>d5</th>\n",
       "      <th>c6</th>\n",
       "      <th>d6</th>\n",
       "      <th>c7</th>\n",
       "      <th>d7</th>\n",
       "      <th>c8</th>\n",
       "      <th>d8</th>\n",
       "      <th>c9</th>\n",
       "      <th>d9</th>\n",
       "      <th>sum_entry</th>\n",
       "      <th>sum_exit</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>seq</th>\n",
       "      <th>traj</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"7\" valign=\"top\">0</th>\n",
       "      <th>0</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.118651</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.0691771</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.0524919</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.0427709</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.0316059</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.0155893</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.00850572</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.00601303</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.00481764</td>\n",
       "      <td>NaN</td>\n",
       "      <td>-0.349623</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>-0.242493</td>\n",
       "      <td>-0.242493</td>\n",
       "      <td>-0.170878</td>\n",
       "      <td>-0.170878</td>\n",
       "      <td>-0.154685</td>\n",
       "      <td>-0.154685</td>\n",
       "      <td>-0.13878</td>\n",
       "      <td>-0.13878</td>\n",
       "      <td>-0.131465</td>\n",
       "      <td>-0.131465</td>\n",
       "      <td>-0.130822</td>\n",
       "      <td>-0.130822</td>\n",
       "      <td>-0.130812</td>\n",
       "      <td>-0.130812</td>\n",
       "      <td>-0.0819298</td>\n",
       "      <td>-0.0819298</td>\n",
       "      <td>-0.0725769</td>\n",
       "      <td>-0.0725769</td>\n",
       "      <td>-0.0574125</td>\n",
       "      <td>-0.0574125</td>\n",
       "      <td>-1.31185</td>\n",
       "      <td>-1.31185</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-0.432722</td>\n",
       "      <td>-0.432722</td>\n",
       "      <td>-0.432425</td>\n",
       "      <td>-0.432425</td>\n",
       "      <td>-0.431771</td>\n",
       "      <td>-0.431771</td>\n",
       "      <td>-0.431658</td>\n",
       "      <td>-0.431658</td>\n",
       "      <td>-0.430884</td>\n",
       "      <td>-0.430884</td>\n",
       "      <td>-0.392217</td>\n",
       "      <td>-0.392217</td>\n",
       "      <td>-0.382518</td>\n",
       "      <td>-0.382518</td>\n",
       "      <td>-0.358415</td>\n",
       "      <td>-0.358415</td>\n",
       "      <td>-0.358167</td>\n",
       "      <td>-0.358167</td>\n",
       "      <td>-0.35735</td>\n",
       "      <td>-0.35735</td>\n",
       "      <td>-4.00813</td>\n",
       "      <td>-4.00813</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>-0.181469</td>\n",
       "      <td>-0.181469</td>\n",
       "      <td>-0.126156</td>\n",
       "      <td>-0.126156</td>\n",
       "      <td>-0.117624</td>\n",
       "      <td>-0.117624</td>\n",
       "      <td>-0.102247</td>\n",
       "      <td>-0.102247</td>\n",
       "      <td>-0.0994271</td>\n",
       "      <td>-0.0994271</td>\n",
       "      <td>-0.0982649</td>\n",
       "      <td>-0.0982649</td>\n",
       "      <td>-0.0782405</td>\n",
       "      <td>-0.0782405</td>\n",
       "      <td>-0.0699517</td>\n",
       "      <td>-0.0699517</td>\n",
       "      <td>-0.0563424</td>\n",
       "      <td>-0.0563424</td>\n",
       "      <td>-0.0499834</td>\n",
       "      <td>-0.0499834</td>\n",
       "      <td>-0.979705</td>\n",
       "      <td>-0.979705</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>-4.04356e-08</td>\n",
       "      <td>-4.04356e-08</td>\n",
       "      <td>-9.31554e-11</td>\n",
       "      <td>-9.31554e-11</td>\n",
       "      <td>-7.83267e-13</td>\n",
       "      <td>-7.83267e-13</td>\n",
       "      <td>-1.09901e-13</td>\n",
       "      <td>-1.09901e-13</td>\n",
       "      <td>-4.46805e-14</td>\n",
       "      <td>-4.46805e-14</td>\n",
       "      <td>-2.23839e-14</td>\n",
       "      <td>-2.23839e-14</td>\n",
       "      <td>-1.8855e-14</td>\n",
       "      <td>-1.8855e-14</td>\n",
       "      <td>-1.05498e-14</td>\n",
       "      <td>-1.05498e-14</td>\n",
       "      <td>-4.80605e-15</td>\n",
       "      <td>-4.80605e-15</td>\n",
       "      <td>-3.57428e-15</td>\n",
       "      <td>-3.57428e-15</td>\n",
       "      <td>-4.05298e-08</td>\n",
       "      <td>-4.05298e-08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>-0.155519</td>\n",
       "      <td>-0.155519</td>\n",
       "      <td>-0.00433187</td>\n",
       "      <td>-0.00433187</td>\n",
       "      <td>-0.000715038</td>\n",
       "      <td>-0.000715038</td>\n",
       "      <td>-0.000117057</td>\n",
       "      <td>-0.000117057</td>\n",
       "      <td>-0.000106357</td>\n",
       "      <td>-0.000106357</td>\n",
       "      <td>-8.59911e-05</td>\n",
       "      <td>-8.59911e-05</td>\n",
       "      <td>-3.17741e-05</td>\n",
       "      <td>-3.17741e-05</td>\n",
       "      <td>-1.24884e-05</td>\n",
       "      <td>-1.24884e-05</td>\n",
       "      <td>-9.53871e-06</td>\n",
       "      <td>-9.53871e-06</td>\n",
       "      <td>-7.45562e-06</td>\n",
       "      <td>-7.45562e-06</td>\n",
       "      <td>-0.160936</td>\n",
       "      <td>-0.160936</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                   c0           d0     ...         sum_entry     sum_exit\n",
       "seq traj                               ...                               \n",
       "0   0             NaN          NaN     ...         -0.349623          NaN\n",
       "    1       -0.242493    -0.242493     ...          -1.31185     -1.31185\n",
       "    2       -0.432722    -0.432722     ...          -4.00813     -4.00813\n",
       "    3       -0.181469    -0.181469     ...         -0.979705    -0.979705\n",
       "    4    -4.04356e-08 -4.04356e-08     ...      -4.05298e-08 -4.05298e-08\n",
       "    5       -0.155519    -0.155519     ...         -0.160936    -0.160936\n",
       "    6             NaN          NaN     ...               NaN          NaN\n",
       "\n",
       "[7 rows x 22 columns]"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_train_cluster_w.head(7)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_train_cluster_w = df_train_cluster_w.drop(columns=df_original_columns)\n",
    "df_test_cluster_w = df_test_cluster_w.drop(columns=df_original_columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_train_cluster_w.to_pickle(\"df_train_cluster_s.pickle\")\n",
    "df_test_cluster_w.to_pickle(\"df_test_cluster_s.pickle\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "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></th>\n",
       "      <th>sum_entry</th>\n",
       "      <th>sum_exit</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>seq</th>\n",
       "      <th>traj</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th rowspan=\"7\" valign=\"top\">0</th>\n",
       "      <th>0</th>\n",
       "      <td>-0.00162238</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>-0.0108394</td>\n",
       "      <td>-0.0108394</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>4.46018</td>\n",
       "      <td>-0.539816</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>-0.0479094</td>\n",
       "      <td>-0.0479094</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>-1.03044</td>\n",
       "      <td>-1.03044</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           sum_entry   sum_exit\n",
       "seq traj                       \n",
       "0   0    -0.00162238        NaN\n",
       "    1     -0.0108394 -0.0108394\n",
       "    2        4.46018  -0.539816\n",
       "    3     -0.0479094 -0.0479094\n",
       "    4       -1.03044   -1.03044\n",
       "    5            NaN        NaN\n",
       "    6            NaN        NaN"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_test_cluster_w.head(7)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}