Regression Modeling.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Regression Modeling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import seaborn as sns\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "\n",
    "import calendar\n",
    "import json\n",
    "from datetime import datetime, timedelta\n",
    "from tqdm import tqdm # progress bar\n",
    "from time import time\n",
    "from collections import Counter\n",
    "\n",
    "import tensorflow as tf\n",
    "from sklearn.metrics import cohen_kappa_score, make_scorer, confusion_matrix\n",
    "from sklearn.model_selection import cross_val_score, KFold, StratifiedKFold, GroupKFold\n",
    "from sklearn.preprocessing import StandardScaler, MinMaxScaler, OneHotEncoder\n",
    "from sklearn.ensemble import RandomForestRegressor\n",
    "from catboost import CatBoostRegressor\n",
    "import xgboost as xgb\n",
    "import lightgbm as lgb\n",
    "from xgboost import XGBRegressor\n",
    "\n",
    "# Ignore warning, should be commented when implementing\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "\n",
    "DATA_PATH = './data'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Loading data\n",
    "train = pd.read_csv(DATA_PATH + './train.csv')\n",
    "specs = pd.read_csv(DATA_PATH + './specs.csv')\n",
    "train_labels = pd.read_csv(DATA_PATH + './train_labels.csv')\n",
    "test = pd.read_csv(DATA_PATH + './test.csv')\n",
    "submission = pd.read_csv(DATA_PATH + './sample_submission.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Encode neccessary features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def encode_title(train, test, train_labels):\n",
    "    # encode title\n",
    "    train['title_event_code'] = list(map(lambda x, y: str(x) + '_' + str(y), train['title'], train['event_code']))\n",
    "    test['title_event_code'] = list(map(lambda x, y: str(x) + '_' + str(y), test['title'], test['event_code']))\n",
    "    all_title_event_code = list(set(train[\"title_event_code\"].unique()).union(test[\"title_event_code\"].unique()))\n",
    "    \n",
    "    # make a list with all the unique 'titles' from the train and test set\n",
    "    list_of_user_activities = list(set(train['title'].unique()).union(set(test['title'].unique())))\n",
    "    \n",
    "    # make a list with all the unique 'event_code' from the train and test set\n",
    "    list_of_event_code = list(set(train['event_code'].unique()).union(set(test['event_code'].unique())))\n",
    "    list_of_event_id = list(set(train['event_id'].unique()).union(set(test['event_id'].unique())))\n",
    "    \n",
    "    # make a list with all the unique worlds from the train and test set\n",
    "    list_of_worlds = list(set(train['world'].unique()).union(set(test['world'].unique())))\n",
    "    \n",
    "    # create a dictionary numerating the titles\n",
    "    activities_map = dict(zip(list_of_user_activities, np.arange(len(list_of_user_activities))))\n",
    "    activities_labels = dict(zip(np.arange(len(list_of_user_activities)), list_of_user_activities))\n",
    "    activities_world = dict(zip(list_of_worlds, np.arange(len(list_of_worlds))))\n",
    "    assess_titles = list(set(train[train['type'] == 'Assessment']['title'].value_counts().index).union(set(test[test['type'] == 'Assessment']['title'].value_counts().index)))\n",
    "   \n",
    "    # replace the text titles with the number titles from the dict\n",
    "    train['title'] = train['title'].map(activities_map)\n",
    "    test['title'] = test['title'].map(activities_map)\n",
    "    train['world'] = train['world'].map(activities_world)\n",
    "    test['world'] = test['world'].map(activities_world)\n",
    "    train_labels['title'] = train_labels['title'].map(activities_map)\n",
    "    win_code = dict(zip(activities_map.values(), (4100*np.ones(len(activities_map))).astype('int')))\n",
    "    \n",
    "    # then, it set one element, the 'Bird Measurer (Assessment)' as 4110, 10 more than the rest\n",
    "    win_code[activities_map['Bird Measurer (Assessment)']] = 4110\n",
    "    \n",
    "    # convert text into datetime\n",
    "    train['timestamp'] = pd.to_datetime(train['timestamp'])\n",
    "    test['timestamp'] = pd.to_datetime(test['timestamp'])\n",
    "    \n",
    "    \n",
    "    return train, test, train_labels, win_code, list_of_user_activities, list_of_event_code, activities_labels, activities_map, assess_titles, list_of_event_id, all_title_event_code"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# get usefull dict with maping encode\n",
    "train, test, train_labels, win_code, list_of_user_activities, list_of_event_code, activities_labels, activities_map, assess_titles,list_of_event_id, all_title_event_code = encode_title(train, test, train_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Clip time dict used for encoding clip duration\n",
    "clip_time = {'Welcome to Lost Lagoon!':19,\n",
    "             'Tree Top City - Level 1':17,\n",
    "             'Ordering Spheres':61, \n",
    "             'Costume Box':61,\n",
    "             '12 Monkeys':109,\n",
    "             'Tree Top City - Level 2':25, \n",
    "             'Pirate\\'s Tale':80, \n",
    "             'Treasure Map':156,\n",
    "             'Tree Top City - Level 3':26,\n",
    "             'Rulers':126, 'Magma Peak - Level 1':20, \n",
    "             'Slop Problem':60, 'Magma Peak - Level 2':22, \n",
    "             'Crystal Caves - Level 1':18,\n",
    "             'Balancing Act':72, 'Lifting Heavy Things':118,\n",
    "             'Crystal Caves - Level 2':24, 'Honey Cake':142, \n",
    "             'Crystal Caves - Level 3':19,\n",
    "             'Heavy, Heavier, Heaviest':61,\n",
    "            }"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "A small notice:\n",
    "> An `installation_id` in **test set** can have many Assessment session but the last Assessment has no attempts (4110 or 4100)\n",
    "\n",
    "### Adding more features to the baseline:\n",
    "- Convert `world` to numerical form\n",
    "- Add corresponding `world` of assessment.\n",
    "    - `assessment_world_code`\n",
    "- Accumulated number of different previous assessments\n",
    "    - `assessment_1`, `assessment_4`, `assessment_16`, `assessment_18`, `assessment_39`, \n",
    "- Add `duration_std` along with `duration_mean`\n",
    "- Add `accumulated_assessment_attemps`\n",
    "- Add combination of `session_title_code` of an assessment with corresponding `accuracy_group`\n",
    "    - `assessment_accuracy_map`\n",
    "- Add clip/activity/game duration mean and std"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_assessment_code(activities_map):\n",
    "    '''\n",
    "    Get code of assessments in activities_map, \n",
    "    since it is randomly generated each time\n",
    "    '''\n",
    "    code_list = []\n",
    "    for key in activities_map.keys():\n",
    "        if 'assessment' in key.lower():\n",
    "            code_list.append(str(activities_map[key]))\n",
    "    return code_list\n",
    "\n",
    "def get_assessment_accuracy_map():\n",
    "    '''\n",
    "    Create a dict mapping accuracy_group to each assessment code\n",
    "    '''\n",
    "    labels = ['0','1','2','3']\n",
    "    assesment_code_list = get_assessment_code(activities_map)\n",
    "    assessment_accuracy_map = {}\n",
    "    for assess_code in assesment_code_list:\n",
    "        for l in labels:\n",
    "            assessment_accuracy_map[assess_code + '_' + l] = 0\n",
    "    del assesment_code_list, labels\n",
    "    return assessment_accuracy_map\n",
    "\n",
    "def get_assessment_code_map():\n",
    "    assesment_code_list = get_assessment_code(activities_map)\n",
    "    assessment_code_map = {}\n",
    "    for assess_code in assesment_code_list:\n",
    "        assessment_code_map['assessment_' + assess_code] = 0\n",
    "    del assesment_code_list\n",
    "    return assessment_code_map\n",
    "\n",
    "def get_data(user_sample, test_set=False):\n",
    "    '''\n",
    "    Take the data grouped by an installation_id and return the assessments with all calculated historic data\n",
    "    '''\n",
    "    assessment_accuracy_map = get_assessment_accuracy_map()\n",
    "    assessment_code_map = get_assessment_code_map()\n",
    "    \n",
    "    user_activities_count = {'type_Clip':0, 'type_Activity': 0, 'type_Assessment': 0, 'type_Game':0, \n",
    "                             'world_0':0, 'world_1':0, 'world_2':0, 'world_3':0}\n",
    "    user_activities_count.update(assessment_code_map)\n",
    "    \n",
    "    # New features updated from sourced kernel\n",
    "    last_accuracy_title = {'acc_' + title: -1 for title in assess_titles}\n",
    "    event_code_count: Dict[str, int] = {ev: 0 for ev in list_of_event_code}\n",
    "    event_id_count: Dict[str, int] = {eve: 0 for eve in list_of_event_id}\n",
    "    title_count: Dict[str, int] = {eve: 0 for eve in activities_labels.values()} \n",
    "    title_event_code_count: Dict[str, int] = {t_eve: 0 for t_eve in all_title_event_code}\n",
    "    \n",
    "    accuracy_groups = {0:0, 1:0, 2:0, 3:0}\n",
    "    all_assessments = []\n",
    "    total_accuracy_group = 0\n",
    "    total_accuracy = 0\n",
    "    accumulated_correct_attempts = 0 \n",
    "    accumulated_uncorrect_attempts = 0 \n",
    "    accumulated_events = 0\n",
    "    assessment_counter = 0\n",
    "    \n",
    "    # Duration features\n",
    "    durations = []\n",
    "    clip_durations = []\n",
    "    activity_durations = []\n",
    "    game_durations = []\n",
    "    \n",
    "    for i, session in user_sample.groupby('game_session', sort=False):\n",
    "        session_type = session['type'].iloc[0]\n",
    "        session_world = session['world'].iloc[0]\n",
    "        session_title = session['title'].iloc[0]\n",
    "        if test_set == True:\n",
    "            second_condition = True\n",
    "        else: # if this is a training set, make sure it has Assessment attemps\n",
    "            if ((session['event_code'] == 4110) | (session['event_code'] == 4100)).any():\n",
    "                second_condition = True\n",
    "            else:\n",
    "                second_condition= False\n",
    "                         \n",
    "        if session_type == 'Clip':\n",
    "            clip_durations.append((clip_time[activities_labels[session_title]]))\n",
    "        \n",
    "        if session_type == 'Activity':\n",
    "            activity_durations.append((session.iloc[-1, 2] - session.iloc[0, 2] ).seconds)\n",
    "        \n",
    "        if session_type == 'Game':\n",
    "            game_durations.append((session.iloc[-1, 2] - session.iloc[0, 2] ).seconds)\n",
    "                \n",
    "        if (session_type == 'Assessment') & (second_condition):\n",
    "            user_activities_count['assessment_'+ str(session_title)] += 1\n",
    "            all_attempts = session.query(f'event_code == {win_code[session_title]}')\n",
    "            true_attempts = all_attempts['event_data'].str.contains('true').sum()\n",
    "            false_attempts = all_attempts['event_data'].str.contains('false').sum()\n",
    "            \n",
    "            features = user_activities_count.copy()\n",
    "            \n",
    "            # New features from the sourced kernel\n",
    "            features.update(last_accuracy_title.copy())\n",
    "            features.update(event_code_count.copy())\n",
    "            features.update(event_id_count.copy())\n",
    "            features.update(title_count.copy())\n",
    "            features.update(title_event_code_count.copy())\n",
    "            variety_features = [('var_event_code', event_code_count),\n",
    "                              ('var_event_id', event_id_count),\n",
    "                               ('var_title', title_count),\n",
    "                               ('var_title_event_code', title_event_code_count)]\n",
    "            \n",
    "            for name, dict_counts in variety_features:\n",
    "                arr = np.array(list(dict_counts.values()))\n",
    "                features[name] = np.count_nonzero(arr)\n",
    "                \n",
    "            features['installation_id'] = session['installation_id'].iloc[0]\n",
    "            features['session_title_code'] = session['title'].iloc[0] \n",
    "            features['assessment_world_code'] = session_world\n",
    "            features['accumulated_correct_attempts'] = accumulated_correct_attempts\n",
    "            features['accumulated_uncorrect_attempts'] = accumulated_uncorrect_attempts\n",
    "            accumulated_correct_attempts += true_attempts \n",
    "            accumulated_uncorrect_attempts += false_attempts\n",
    "            if durations == []:\n",
    "                features['duration_mean'] = 0\n",
    "                features['duration_std'] = 0\n",
    "            else:\n",
    "                features['duration_mean'] = np.mean(durations)\n",
    "                features['duration_std'] = np.std(durations)\n",
    "            durations.append((session.iloc[-1]['timestamp'] - session.iloc[0]['timestamp']).seconds)\n",
    "            \n",
    "            if clip_durations == []:\n",
    "                features['Clip_duration_mean'] = 0\n",
    "                features['Clip_duration_std'] = 0\n",
    "            else:\n",
    "                features['Clip_duration_mean'] = np.mean(clip_durations)\n",
    "                features['Clip_duration_std'] = np.std(clip_durations)\n",
    "                \n",
    "            if activity_durations == []:\n",
    "                features['Activity_duration_mean'] = 0\n",
    "                features['Activity_duration_std'] = 0\n",
    "            else:\n",
    "                features['Activity_duration_mean'] = np.mean(activity_durations)\n",
    "                features['Activity_duration_std'] = np.std(activity_durations)\n",
    "                \n",
    "            if game_durations == []:\n",
    "                features['Game_duration_mean'] = 0\n",
    "                features['Game_duration_std'] = 0\n",
    "            else:\n",
    "                features['Game_duration_mean'] = np.mean(game_durations)\n",
    "                features['Game_duration_std'] = np.std(game_durations)\n",
    "            \n",
    "            features['accumulated_accuracy'] = total_accuracy/assessment_counter if assessment_counter > 0 else 0\n",
    "            accuracy = true_attempts/(true_attempts+false_attempts) if (true_attempts+false_attempts) != 0 else 0\n",
    "            total_accuracy += accuracy\n",
    "            if accuracy == 0:\n",
    "                features['accuracy_group'] = 0\n",
    "            elif accuracy == 1:\n",
    "                features['accuracy_group'] = 3\n",
    "            elif accuracy == 0.5:\n",
    "                features['accuracy_group'] = 2\n",
    "            else:\n",
    "                features['accuracy_group'] = 1\n",
    "                \n",
    "            features.update(accuracy_groups)\n",
    "            features.update(assessment_accuracy_map)\n",
    "            \n",
    "            features['accumulated_assessment_attemps'] = sum(accuracy_groups.values())\n",
    "            features['accumulated_accuracy_group'] = total_accuracy_group/assessment_counter if assessment_counter > 0 else 0\n",
    "            features['accumulated_events'] = accumulated_events\n",
    "            total_accuracy_group += features['accuracy_group']\n",
    "            accuracy_groups[features['accuracy_group']] += 1\n",
    "            assessment_accuracy_map[str(session['title'].iloc[0]) + '_' + str(features['accuracy_group'])] += 1\n",
    "            \n",
    "            all_assessments.append(features)\n",
    "            assessment_counter += 1\n",
    "            \n",
    "        # this piece counts how many actions were made in each event_code so far\n",
    "        def update_counters(counter: dict, col: str):\n",
    "            num_of_session_count = Counter(session[col])\n",
    "            for k in num_of_session_count.keys():\n",
    "                x = k\n",
    "                if col == 'title':\n",
    "                    x = activities_labels[k]\n",
    "                counter[x] += num_of_session_count[k]\n",
    "            return counter\n",
    "            \n",
    "        event_code_count = update_counters(event_code_count, \"event_code\")\n",
    "        event_id_count = update_counters(event_id_count, \"event_id\")\n",
    "        title_count = update_counters(title_count, 'title')\n",
    "        title_event_code_count = update_counters(title_event_code_count, 'title_event_code')\n",
    "        \n",
    "        \n",
    "        accumulated_events += len(session)\n",
    "        user_activities_count['type_'+session_type] += 1\n",
    "        user_activities_count['world_'+str(session_world)] += 1\n",
    "\n",
    "    if test_set:\n",
    "        return all_assessments[-1]\n",
    "    return all_assessments"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████| 17000/17000 [07:23<00:00, 38.35it/s]\n"
     ]
    }
   ],
   "source": [
    "compiled_train = []\n",
    "for i, (ins_id, user_sample) in tqdm(enumerate(train.groupby('installation_id', sort=False)), total=17000):\n",
    "    compiled_train += get_data(user_sample)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(17692, 937)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "new_train = pd.DataFrame(compiled_train)\n",
    "del compiled_train\n",
    "new_train.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Processing Test set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████████████████| 1000/1000 [00:27<00:00, 36.87it/s]\n"
     ]
    }
   ],
   "source": [
    "# Processing test set\n",
    "compiled_test = []\n",
    "for ins_id, user_sample in tqdm(test.groupby('installation_id', sort=False), total=1000):\n",
    "    a = get_data(user_sample, test_set=True)\n",
    "    compiled_test.append(a)\n",
    "    \n",
    "new_test = pd.DataFrame(compiled_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature Selection\n",
    "\n",
    "- Eliminate high correlated features ? Doing this seems to lower the score of public test set.\n",
    "- Scaling features:\n",
    "    - Standardizing would lead to a very low score of test set. \n",
    "    - Furthermore, tree-based models (lightgbm, xgboost, catboost) are not distance based models and hence, can handle varying range of features.\n",
    "- Eliminate low-variance features?\n",
    "    - There seems to be no low-variance feature"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Preprocess feature names to avoid \"Special JSON characters\" error of LightGBM\n",
    "new_train.columns = [\"\".join (c if c.isalnum() else \"_\" for c in str(x)) for x in new_train.columns]\n",
    "new_test.columns = [\"\".join (c if c.isalnum() else \"_\" for c in str(x)) for x in new_test.columns]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Call feature engineering functions\n",
    "cat_features = ['session_title_code','assessment_world_code']\n",
    "features = new_train.loc[(new_train.sum(axis=1) != 0), (new_train.sum(axis=0) != 0)].columns # delete useless columns\n",
    "features = [x for x in features if x not in ['accuracy_group','installation_id']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1: FEAT_A: type_Clip FEAT_B: 27253bdc - Correlation: 1.0\n",
      "2: FEAT_A: world_3 FEAT_B: Welcome_to_Lost_Lagoon_ - Correlation: 1.0\n",
      "3: FEAT_A: world_3 FEAT_B: Welcome_to_Lost_Lagoon__2000 - Correlation: 1.0\n",
      "4: FEAT_A: assessment_20 FEAT_B: 9d29771f - Correlation: 0.9902367941418727\n",
      "5: FEAT_A: assessment_20 FEAT_B: c74f40cd - Correlation: 0.9902297744191059\n",
      "6: FEAT_A: assessment_20 FEAT_B: Mushroom_Sorter__Assessment__3021 - Correlation: 0.9902367941418727\n",
      "7: FEAT_A: assessment_20 FEAT_B: Mushroom_Sorter__Assessment__3121 - Correlation: 0.9902297744191059\n",
      "8: FEAT_A: 2050 FEAT_B: 2040 - Correlation: 0.9965259852533822\n",
      "9: FEAT_A: 2050 FEAT_B: 08fd73f3 - Correlation: 0.9966124346503646\n",
      "10: FEAT_A: 2050 FEAT_B: f71c4741 - Correlation: 0.9904124425193319\n",
      "11: FEAT_A: 2050 FEAT_B: 5a848010 - Correlation: 0.992355236510153\n",
      "12: FEAT_A: 2050 FEAT_B: 37c53127 - Correlation: 1.0\n",
      "13: FEAT_A: 2050 FEAT_B: 2b9272f4 - Correlation: 0.999983903256798\n",
      "14: FEAT_A: 2050 FEAT_B: f7e47413 - Correlation: 0.9902188488377782\n",
      "15: FEAT_A: 2050 FEAT_B: 73757a5e - Correlation: 0.9998050177953759\n",
      "16: FEAT_A: 2050 FEAT_B: 26fd2d99 - Correlation: 0.9965085078132824\n",
      "17: FEAT_A: 2050 FEAT_B: dcaede90 - Correlation: 0.9965259852533822\n",
      "18: FEAT_A: 2050 FEAT_B: Scrub_A_Dub_2020 - Correlation: 0.9965085078132824\n",
      "19: FEAT_A: 2050 FEAT_B: Scrub_A_Dub_2040 - Correlation: 0.9965259852533822\n",
      "20: FEAT_A: 2050 FEAT_B: Scrub_A_Dub_2030 - Correlation: 0.9966124346503646\n",
      "21: FEAT_A: 2050 FEAT_B: Scrub_A_Dub_3110 - Correlation: 0.9902188488377782\n",
      "22: FEAT_A: 2050 FEAT_B: Scrub_A_Dub_2080 - Correlation: 0.992355236510153\n",
      "23: FEAT_A: 2050 FEAT_B: Scrub_A_Dub_3121 - Correlation: 0.999983903256798\n",
      "24: FEAT_A: 2050 FEAT_B: Scrub_A_Dub_3010 - Correlation: 0.9904124425193319\n",
      "25: FEAT_A: 2050 FEAT_B: Scrub_A_Dub_3021 - Correlation: 0.9998050177953759\n",
      "26: FEAT_A: 2050 FEAT_B: Scrub_A_Dub_2050 - Correlation: 1.0\n",
      "27: FEAT_A: 4110 FEAT_B: 17113b36 - Correlation: 0.9921889902945864\n",
      "28: FEAT_A: 4110 FEAT_B: e37a2b78 - Correlation: 0.9916266087483759\n",
      "29: FEAT_A: 4110 FEAT_B: Bird_Measurer__Assessment__3120 - Correlation: 0.9916266087483759\n",
      "30: FEAT_A: 4110 FEAT_B: Bird_Measurer__Assessment__4110 - Correlation: 0.9921889902945864\n",
      "31: FEAT_A: 3110 FEAT_B: 3010 - Correlation: 0.9999293423389526\n",
      "32: FEAT_A: 3120 FEAT_B: 3020 - Correlation: 0.9998761454217059\n",
      "33: FEAT_A: 3121 FEAT_B: 3021 - Correlation: 0.9999098212778367\n",
      "34: FEAT_A: 4220 FEAT_B: 1340b8d7 - Correlation: 1.0\n",
      "35: FEAT_A: 4220 FEAT_B: Bubble_Bath_4220 - Correlation: 1.0\n",
      "36: FEAT_A: 4230 FEAT_B: 4235 - Correlation: 0.9999995197529113\n",
      "37: FEAT_A: 4230 FEAT_B: ad148f58 - Correlation: 0.9999999999999998\n",
      "38: FEAT_A: 4230 FEAT_B: 85de926c - Correlation: 0.9999995197529113\n",
      "39: FEAT_A: 4230 FEAT_B: Bubble_Bath_4230 - Correlation: 0.9999999999999998\n",
      "40: FEAT_A: 4230 FEAT_B: Bubble_Bath_4235 - Correlation: 0.9999995197529113\n",
      "41: FEAT_A: 5000 FEAT_B: 5010 - Correlation: 0.99918493999936\n",
      "42: FEAT_A: 5000 FEAT_B: 71e712d8 - Correlation: 0.99918493999936\n",
      "43: FEAT_A: 5000 FEAT_B: a6d66e51 - Correlation: 1.0\n",
      "44: FEAT_A: 5000 FEAT_B: Watering_Hole__Activity__5000 - Correlation: 1.0\n",
      "45: FEAT_A: 5000 FEAT_B: Watering_Hole__Activity__5010 - Correlation: 0.99918493999936\n",
      "46: FEAT_A: 4031 FEAT_B: 1996c610 - Correlation: 1.0\n",
      "47: FEAT_A: 4031 FEAT_B: 74e5f8a7 - Correlation: 0.992500794026255\n",
      "48: FEAT_A: 4031 FEAT_B: Dino_Drink_4020 - Correlation: 0.992500794026255\n",
      "49: FEAT_A: 4031 FEAT_B: Dino_Drink_4031 - Correlation: 1.0\n",
      "50: FEAT_A: 4050 FEAT_B: a1192f43 - Correlation: 1.0\n",
      "51: FEAT_A: 4050 FEAT_B: Crystals_Rule_4050 - Correlation: 1.0\n",
      "52: FEAT_A: 2020 FEAT_B: 2030 - Correlation: 0.9959934301436908\n",
      "53: FEAT_A: 7ad3efc6 FEAT_B: 65a38bf7 - Correlation: 0.9999786271134931\n",
      "54: FEAT_A: 7ad3efc6 FEAT_B: Cart_Balancer__Assessment__2000 - Correlation: 1.0\n",
      "55: FEAT_A: 7ad3efc6 FEAT_B: Cart_Balancer__Assessment__2020 - Correlation: 0.9999786271134931\n",
      "56: FEAT_A: 90d848e0 FEAT_B: Cauldron_Filler__Assessment__2000 - Correlation: 1.0\n",
      "57: FEAT_A: 15a43e5b FEAT_B: Bottle_Filler__Activity__4070 - Correlation: 0.9999999999999999\n",
      "58: FEAT_A: 86c924c4 FEAT_B: 3323d7e9 - Correlation: 0.9930333069192077\n",
      "59: FEAT_A: 86c924c4 FEAT_B: 7cf1bc53 - Correlation: 0.9943504919785926\n",
      "60: FEAT_A: 86c924c4 FEAT_B: 3ddc79c3 - Correlation: 0.9923073698915048\n",
      "61: FEAT_A: 86c924c4 FEAT_B: 3babcb9b - Correlation: 0.9988970597854074\n",
      "62: FEAT_A: 86c924c4 FEAT_B: e720d930 - Correlation: 0.9921247688381784\n",
      "63: FEAT_A: 86c924c4 FEAT_B: 5154fc30 - Correlation: 0.9988936438976114\n",
      "64: FEAT_A: 86c924c4 FEAT_B: Crystals_Rule_3110 - Correlation: 0.9988970597854074\n",
      "65: FEAT_A: 86c924c4 FEAT_B: Crystals_Rule_3021 - Correlation: 0.9923073698915048\n",
      "66: FEAT_A: 86c924c4 FEAT_B: Crystals_Rule_2030 - Correlation: 0.9930333069192077\n",
      "67: FEAT_A: 86c924c4 FEAT_B: Crystals_Rule_3121 - Correlation: 0.9921247688381784\n",
      "68: FEAT_A: 86c924c4 FEAT_B: Crystals_Rule_2020 - Correlation: 0.9943504919785926\n",
      "69: FEAT_A: 86c924c4 FEAT_B: Crystals_Rule_4020 - Correlation: 0.9999999999999999\n",
      "70: FEAT_A: 86c924c4 FEAT_B: Crystals_Rule_3010 - Correlation: 0.9988936438976114\n",
      "71: FEAT_A: 5c2f29ca FEAT_B: Cart_Balancer__Assessment__4020 - Correlation: 1.0\n",
      "72: FEAT_A: c51d8688 FEAT_B: 907a054b - Correlation: 0.9999667373226213\n",
      "73: FEAT_A: c51d8688 FEAT_B: Pan_Balance_3020 - Correlation: 0.9999667373226213\n",
      "74: FEAT_A: c51d8688 FEAT_B: Pan_Balance_3120 - Correlation: 1.0\n",
      "75: FEAT_A: 7ec0c298 FEAT_B: Chow_Time_3010 - Correlation: 0.9999999999999999\n",
      "76: FEAT_A: f806dc10 FEAT_B: Dino_Drink_2020 - Correlation: 1.0\n",
      "77: FEAT_A: d06f75b5 FEAT_B: c54cf6c5 - Correlation: 0.997238403362544\n",
      "78: FEAT_A: d06f75b5 FEAT_B: 3bb91dda - Correlation: 0.9970877890329345\n",
      "79: FEAT_A: d06f75b5 FEAT_B: 8f094001 - Correlation: 0.9925750877376139\n",
      "80: FEAT_A: d06f75b5 FEAT_B: 15eb4a7d - Correlation: 0.9907849874686498\n",
      "81: FEAT_A: d06f75b5 FEAT_B: 1beb320a - Correlation: 0.9906445820168198\n",
      "82: FEAT_A: d06f75b5 FEAT_B: 895865f3 - Correlation: 0.9991496587153846\n",
      "83: FEAT_A: d06f75b5 FEAT_B: Bubble_Bath_2035 - Correlation: 1.0\n",
      "84: FEAT_A: d06f75b5 FEAT_B: Bubble_Bath_4020 - Correlation: 0.9970877890329345\n",
      "85: FEAT_A: d06f75b5 FEAT_B: Bubble_Bath_4045 - Correlation: 0.9925750877376139\n",
      "86: FEAT_A: d06f75b5 FEAT_B: Bubble_Bath_2020 - Correlation: 0.9906445820168198\n",
      "87: FEAT_A: d06f75b5 FEAT_B: Bubble_Bath_2025 - Correlation: 0.997238403362544\n",
      "88: FEAT_A: d06f75b5 FEAT_B: Bubble_Bath_3110 - Correlation: 0.9907849874686498\n",
      "89: FEAT_A: d06f75b5 FEAT_B: Bubble_Bath_2030 - Correlation: 0.9991496587153846\n",
      "90: FEAT_A: 69fdac0a FEAT_B: 8d7e386c - Correlation: 0.9996590304051964\n",
      "91: FEAT_A: 69fdac0a FEAT_B: Happy_Camel_3010 - Correlation: 0.9996590304051964\n",
      "92: FEAT_A: 69fdac0a FEAT_B: Happy_Camel_3110 - Correlation: 0.9999999999999999\n",
      "93: FEAT_A: bcceccc6 FEAT_B: Air_Show_4070 - Correlation: 1.0\n",
      "94: FEAT_A: 1375ccb7 FEAT_B: bdf49a58 - Correlation: 0.9993801932996246\n",
      "95: FEAT_A: 1375ccb7 FEAT_B: Bird_Measurer__Assessment__3110 - Correlation: 0.9993801932996246\n",
      "96: FEAT_A: 1375ccb7 FEAT_B: Bird_Measurer__Assessment__3010 - Correlation: 1.0\n",
      "97: FEAT_A: e080a381 FEAT_B: Pan_Balance_4090 - Correlation: 1.0\n",
      "98: FEAT_A: 5d042115 FEAT_B: Flower_Waterer__Activity__4030 - Correlation: 1.0\n",
      "99: FEAT_A: a2df0760 FEAT_B: Happy_Camel_4035 - Correlation: 1.0\n",
      "100: FEAT_A: 9de5e594 FEAT_B: 28a4eb9a - Correlation: 0.9995923610403482\n",
      "101: FEAT_A: 9de5e594 FEAT_B: Dino_Dive_3020 - Correlation: 1.0\n",
      "102: FEAT_A: 9de5e594 FEAT_B: Dino_Dive_3120 - Correlation: 0.9995923610403482\n",
      "103: FEAT_A: e5734469 FEAT_B: 89aace00 - Correlation: 0.9998406136655407\n",
      "104: FEAT_A: e5734469 FEAT_B: Dino_Drink_3020 - Correlation: 1.0\n",
      "105: FEAT_A: e5734469 FEAT_B: Dino_Drink_3120 - Correlation: 0.9998406136655407\n",
      "106: FEAT_A: 8d748b58 FEAT_B: Bug_Measurer__Activity__4090 - Correlation: 1.0\n",
      "107: FEAT_A: 3d8c61b0 FEAT_B: Happy_Camel - Correlation: 0.9926346677355283\n",
      "108: FEAT_A: 3d8c61b0 FEAT_B: Happy_Camel_4030 - Correlation: 1.0\n",
      "109: FEAT_A: cf7638f3 FEAT_B: a592d54e - Correlation: 0.9969813657046029\n",
      "110: FEAT_A: cf7638f3 FEAT_B: 250513af - Correlation: 0.9996303674081681\n",
      "111: FEAT_A: cf7638f3 FEAT_B: 1c178d24 - Correlation: 0.9996196429168017\n",
      "112: FEAT_A: cf7638f3 FEAT_B: Pan_Balance_2020 - Correlation: 0.9969813657046029\n",
      "113: FEAT_A: cf7638f3 FEAT_B: Pan_Balance_2030 - Correlation: 0.9996196429168017\n",
      "114: FEAT_A: cf7638f3 FEAT_B: Pan_Balance_3121 - Correlation: 1.0\n",
      "115: FEAT_A: cf7638f3 FEAT_B: Pan_Balance_3021 - Correlation: 0.9996303674081681\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "116: FEAT_A: 736f9581 FEAT_B: 9b23e8ee - Correlation: 1.0\n",
      "117: FEAT_A: 736f9581 FEAT_B: Egg_Dropper__Activity__2020 - Correlation: 1.0\n",
      "118: FEAT_A: 736f9581 FEAT_B: Egg_Dropper__Activity__2000 - Correlation: 1.0\n",
      "119: FEAT_A: 5f5b2617 FEAT_B: Bottle_Filler__Activity__4080 - Correlation: 0.9999999999999999\n",
      "120: FEAT_A: 7d5c30a2 FEAT_B: Dino_Dive_2060 - Correlation: 1.0\n",
      "121: FEAT_A: bc8f2793 FEAT_B: Pan_Balance_4035 - Correlation: 1.0\n",
      "122: FEAT_A: 29bdd9ba FEAT_B: Dino_Dive_2000 - Correlation: 1.0\n",
      "123: FEAT_A: 3bf1cf26 FEAT_B: 1af8be29 - Correlation: 0.9998900860839074\n",
      "124: FEAT_A: 3bf1cf26 FEAT_B: Happy_Camel_3020 - Correlation: 0.9998900860839074\n",
      "125: FEAT_A: 3bf1cf26 FEAT_B: Happy_Camel_3120 - Correlation: 0.9999999999999999\n",
      "126: FEAT_A: 88d4a5be FEAT_B: 160654fd - Correlation: 0.9989297106527708\n",
      "127: FEAT_A: 88d4a5be FEAT_B: Mushroom_Sorter__Assessment__3020 - Correlation: 0.9989297106527708\n",
      "128: FEAT_A: 88d4a5be FEAT_B: Mushroom_Sorter__Assessment__3120 - Correlation: 1.0\n",
      "129: FEAT_A: a76029ee FEAT_B: Bird_Measurer__Assessment__4040 - Correlation: 1.0\n",
      "130: FEAT_A: ab3136ba FEAT_B: 832735e1 - Correlation: 0.9998637973165293\n",
      "131: FEAT_A: ab3136ba FEAT_B: Dino_Dive_3010 - Correlation: 0.9998637973165293\n",
      "132: FEAT_A: ab3136ba FEAT_B: Dino_Dive_3110 - Correlation: 1.0\n",
      "133: FEAT_A: 4d6737eb FEAT_B: Dino_Drink_2070 - Correlation: 1.0\n",
      "134: FEAT_A: a5be6304 FEAT_B: 6c930e6e - Correlation: 0.9962760879456208\n",
      "135: FEAT_A: a5be6304 FEAT_B: Mushroom_Sorter__Assessment__2030 - Correlation: 0.9962760879456208\n",
      "136: FEAT_A: a5be6304 FEAT_B: Mushroom_Sorter__Assessment__2010 - Correlation: 1.0\n",
      "137: FEAT_A: 05ad839b FEAT_B: Happy_Camel_4090 - Correlation: 0.9999999999999999\n",
      "138: FEAT_A: 7040c096 FEAT_B: Scrub_A_Dub_4010 - Correlation: 1.0\n",
      "139: FEAT_A: 0a08139c FEAT_B: 71fe8f75 - Correlation: 0.999985034580334\n",
      "140: FEAT_A: 0a08139c FEAT_B: Bug_Measurer__Activity__3010 - Correlation: 1.0\n",
      "141: FEAT_A: 0a08139c FEAT_B: Bug_Measurer__Activity__3110 - Correlation: 0.999985034580334\n",
      "142: FEAT_A: 7dfe6d8a FEAT_B: Leaf_Leader_4070 - Correlation: 0.9999999999999998\n",
      "143: FEAT_A: ecc6157f FEAT_B: Cart_Balancer__Assessment__4080 - Correlation: 1.0\n",
      "144: FEAT_A: 9b01374f FEAT_B: Flower_Waterer__Activity__2000 - Correlation: 1.0\n",
      "145: FEAT_A: 2230fab4 FEAT_B: cfbd47c8 - Correlation: 0.9929107933232209\n",
      "146: FEAT_A: 2230fab4 FEAT_B: 0330ab6a - Correlation: 0.9998673402320392\n",
      "147: FEAT_A: 2230fab4 FEAT_B: Chow_Time_3020 - Correlation: 0.9998673402320392\n",
      "148: FEAT_A: 2230fab4 FEAT_B: Chow_Time_3120 - Correlation: 1.0\n",
      "149: FEAT_A: 2230fab4 FEAT_B: Chow_Time_4030 - Correlation: 0.9929107933232209\n",
      "150: FEAT_A: 3bb91ced FEAT_B: Happy_Camel_2081 - Correlation: 1.0\n",
      "151: FEAT_A: b2e5b0f1 FEAT_B: b74258a0 - Correlation: 0.9998494703471165\n",
      "152: FEAT_A: b2e5b0f1 FEAT_B: ecaab346 - Correlation: 0.9998494703471165\n",
      "153: FEAT_A: b2e5b0f1 FEAT_B: Cart_Balancer__Assessment__3121 - Correlation: 0.9998494703471165\n",
      "154: FEAT_A: b2e5b0f1 FEAT_B: Cart_Balancer__Assessment__2010 - Correlation: 1.0\n",
      "155: FEAT_A: b2e5b0f1 FEAT_B: Cart_Balancer__Assessment__2030 - Correlation: 0.9998494703471165\n",
      "156: FEAT_A: df4940d3 FEAT_B: 90efca10 - Correlation: 0.9909360649697788\n",
      "157: FEAT_A: df4940d3 FEAT_B: b7530680 - Correlation: 0.9939209403211885\n",
      "158: FEAT_A: df4940d3 FEAT_B: e9c52111 - Correlation: 0.9902486005398765\n",
      "159: FEAT_A: df4940d3 FEAT_B: 67439901 - Correlation: 0.9999351649172006\n",
      "160: FEAT_A: df4940d3 FEAT_B: Bottle_Filler__Activity__2020 - Correlation: 0.9939209403211885\n",
      "161: FEAT_A: df4940d3 FEAT_B: Bottle_Filler__Activity__3110 - Correlation: 1.0\n",
      "162: FEAT_A: df4940d3 FEAT_B: Bottle_Filler__Activity__3010 - Correlation: 0.9999351649172006\n",
      "163: FEAT_A: df4940d3 FEAT_B: Bottle_Filler__Activity__4020 - Correlation: 0.9909360649697788\n",
      "164: FEAT_A: df4940d3 FEAT_B: Bottle_Filler__Activity__2030 - Correlation: 0.9902486005398765\n",
      "165: FEAT_A: 77261ab5 FEAT_B: Sandcastle_Builder__Activity__2000 - Correlation: 0.9999999999999999\n",
      "166: FEAT_A: 93edfe2e FEAT_B: Crystals_Rule_4090 - Correlation: 1.0\n",
      "167: FEAT_A: 58a0de5c FEAT_B: 9b4001e4 - Correlation: 0.9999159370503793\n",
      "168: FEAT_A: 58a0de5c FEAT_B: f5b8c21a - Correlation: 0.9977888317691139\n",
      "169: FEAT_A: 58a0de5c FEAT_B: 1575e76c - Correlation: 0.992474858875488\n",
      "170: FEAT_A: 58a0de5c FEAT_B: Air_Show_3121 - Correlation: 1.0\n",
      "171: FEAT_A: 58a0de5c FEAT_B: Air_Show_2030 - Correlation: 0.9977888317691139\n",
      "172: FEAT_A: 58a0de5c FEAT_B: Air_Show_3021 - Correlation: 0.9999159370503793\n",
      "173: FEAT_A: 58a0de5c FEAT_B: Air_Show_2020 - Correlation: 0.992474858875488\n",
      "174: FEAT_A: 16667cc5 FEAT_B: Chicken_Balancer__Activity__4080 - Correlation: 1.0\n",
      "175: FEAT_A: ecc36b7f FEAT_B: Bubble_Bath_4095 - Correlation: 0.9999999999999998\n",
      "176: FEAT_A: 3d63345e FEAT_B: Cart_Balancer__Assessment__4035 - Correlation: 1.0\n",
      "177: FEAT_A: 38074c54 FEAT_B: 222660ff - Correlation: 1.0\n",
      "178: FEAT_A: 38074c54 FEAT_B: e4f1efe6 - Correlation: 0.9984044891339933\n",
      "179: FEAT_A: 38074c54 FEAT_B: 3afb49e6 - Correlation: 0.9939318286621426\n",
      "180: FEAT_A: 38074c54 FEAT_B: Chest_Sorter__Assessment__2010 - Correlation: 1.0\n",
      "181: FEAT_A: 38074c54 FEAT_B: Chest_Sorter__Assessment__2030 - Correlation: 1.0\n",
      "182: FEAT_A: 38074c54 FEAT_B: Chest_Sorter__Assessment__3021 - Correlation: 0.9939318286621426\n",
      "183: FEAT_A: 38074c54 FEAT_B: Chest_Sorter__Assessment__3121 - Correlation: 0.9984044891339933\n",
      "184: FEAT_A: fcfdffb6 FEAT_B: Flower_Waterer__Activity__4022 - Correlation: 1.0\n",
      "185: FEAT_A: a0faea5d FEAT_B: Bubble_Bath_4070 - Correlation: 1.0\n",
      "186: FEAT_A: 28ed704e FEAT_B: 3dfd4aa4 - Correlation: 0.9996618557199024\n",
      "187: FEAT_A: 28ed704e FEAT_B: a52b92d5 - Correlation: 0.9913417216759441\n",
      "188: FEAT_A: 28ed704e FEAT_B: 83c6c409 - Correlation: 0.9996646826146621\n",
      "189: FEAT_A: 28ed704e FEAT_B: Mushroom_Sorter__Assessment__4025 - Correlation: 1.0\n",
      "190: FEAT_A: 28ed704e FEAT_B: Mushroom_Sorter__Assessment__2035 - Correlation: 0.9996646826146621\n",
      "191: FEAT_A: 28ed704e FEAT_B: Mushroom_Sorter__Assessment__2020 - Correlation: 0.9996618557199024\n",
      "192: FEAT_A: 28ed704e FEAT_B: Mushroom_Sorter__Assessment__3110 - Correlation: 0.9913417216759441\n",
      "193: FEAT_A: e79f3763 FEAT_B: Bug_Measurer__Activity__4030 - Correlation: 1.0\n",
      "194: FEAT_A: f56e0afc FEAT_B: Bird_Measurer__Assessment__2000 - Correlation: 0.9999999999999998\n",
      "195: FEAT_A: 7961e599 FEAT_B: 709b1251 - Correlation: 0.9950005827911984\n",
      "196: FEAT_A: 7961e599 FEAT_B: 00c73085 - Correlation: 0.9924736420118322\n",
      "197: FEAT_A: 7961e599 FEAT_B: e3ff61fb - Correlation: 0.9941263988935392\n",
      "198: FEAT_A: 7961e599 FEAT_B: Dino_Dive_2030 - Correlation: 0.9924736420118322\n",
      "199: FEAT_A: 7961e599 FEAT_B: Dino_Dive_3121 - Correlation: 0.9950005827911984\n",
      "200: FEAT_A: 7961e599 FEAT_B: Dino_Dive_2020 - Correlation: 1.0\n",
      "201: FEAT_A: 7961e599 FEAT_B: Dino_Dive_3021 - Correlation: 0.9941263988935392\n",
      "202: FEAT_A: c952eb01 FEAT_B: Watering_Hole__Activity__4070 - Correlation: 0.9999999999999998\n",
      "203: FEAT_A: 02a42007 FEAT_B: Fireworks__Activity__4030 - Correlation: 1.0\n",
      "204: FEAT_A: 2dc29e21 FEAT_B: All_Star_Sorting_4020 - Correlation: 1.0\n",
      "205: FEAT_A: 363c86c9 FEAT_B: Bug_Measurer__Activity__4035 - Correlation: 1.0\n",
      "206: FEAT_A: 5e109ec3 FEAT_B: Cart_Balancer__Assessment__4030 - Correlation: 1.0\n",
      "207: FEAT_A: 7525289a FEAT_B: 45d01abe - Correlation: 0.9981555554467466\n",
      "208: FEAT_A: 7525289a FEAT_B: Bird_Measurer__Assessment__3021 - Correlation: 0.9981555554467466\n",
      "209: FEAT_A: 7525289a FEAT_B: Bird_Measurer__Assessment__3121 - Correlation: 0.9999999999999998\n",
      "210: FEAT_A: 8af75982 FEAT_B: Happy_Camel_4020 - Correlation: 1.0\n",
      "211: FEAT_A: 9e34ea74 FEAT_B: Egg_Dropper__Activity__4070 - Correlation: 1.0\n",
      "212: FEAT_A: 2c4e6db0 FEAT_B: All_Star_Sorting_2020 - Correlation: 1.0\n",
      "213: FEAT_A: 2fb91ec1 FEAT_B: d2e9262e - Correlation: 0.999143468104081\n",
      "214: FEAT_A: 2fb91ec1 FEAT_B: Watering_Hole__Activity__4020 - Correlation: 0.999143468104081\n",
      "215: FEAT_A: 2fb91ec1 FEAT_B: Watering_Hole__Activity__4025 - Correlation: 0.9999999999999998\n",
      "216: FEAT_A: 0d18d96c FEAT_B: Mushroom_Sorter__Assessment__4035 - Correlation: 0.9999999999999999\n",
      "217: FEAT_A: 2a512369 FEAT_B: 33505eae - Correlation: 0.9994585290234326\n",
      "218: FEAT_A: 2a512369 FEAT_B: Leaf_Leader_3110 - Correlation: 1.0\n",
      "219: FEAT_A: 2a512369 FEAT_B: Leaf_Leader_3010 - Correlation: 0.9994585290234326\n",
      "220: FEAT_A: 31973d56 FEAT_B: 5de79a6a - Correlation: 0.9973945622656409\n",
      "221: FEAT_A: 31973d56 FEAT_B: Cart_Balancer__Assessment__3020 - Correlation: 0.9973945622656409\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "222: FEAT_A: 31973d56 FEAT_B: Cart_Balancer__Assessment__3120 - Correlation: 1.0\n",
      "223: FEAT_A: db02c830 FEAT_B: 3bfd1a65 - Correlation: 0.9999982205419902\n",
      "224: FEAT_A: db02c830 FEAT_B: Mushroom_Sorter__Assessment__2025 - Correlation: 0.9999999999999999\n",
      "225: FEAT_A: db02c830 FEAT_B: Mushroom_Sorter__Assessment__2000 - Correlation: 0.9999982205419902\n",
      "226: FEAT_A: 5e3ea25a FEAT_B: Crystals_Rule_4070 - Correlation: 1.0\n",
      "227: FEAT_A: 4e5fc6f5 FEAT_B: Cart_Balancer__Assessment__4090 - Correlation: 1.0\n",
      "228: FEAT_A: c6971acf FEAT_B: 51311d7a - Correlation: 0.9931667723288344\n",
      "229: FEAT_A: c6971acf FEAT_B: Dino_Drink_2000 - Correlation: 0.9931667723288344\n",
      "230: FEAT_A: c6971acf FEAT_B: Dino_Drink_2060 - Correlation: 1.0\n",
      "231: FEAT_A: 1cc7cfca FEAT_B: All_Star_Sorting_4030 - Correlation: 1.0\n",
      "232: FEAT_A: 3edf6747 FEAT_B: Cauldron_Filler__Assessment__4035 - Correlation: 1.0\n",
      "233: FEAT_A: 6d90d394 FEAT_B: Scrub_A_Dub_2000 - Correlation: 1.0\n",
      "234: FEAT_A: e694a35b FEAT_B: Fireworks__Activity__4020 - Correlation: 1.0\n",
      "235: FEAT_A: b88f38da FEAT_B: beb0a7b9 - Correlation: 0.999912519860299\n",
      "236: FEAT_A: b88f38da FEAT_B: Fireworks__Activity__3110 - Correlation: 1.0\n",
      "237: FEAT_A: b88f38da FEAT_B: Fireworks__Activity__3010 - Correlation: 0.999912519860299\n",
      "238: FEAT_A: 30614231 FEAT_B: 2dcad279 - Correlation: 0.9914755860152119\n",
      "239: FEAT_A: 30614231 FEAT_B: 923afab1 - Correlation: 0.9915794451958683\n",
      "240: FEAT_A: 30614231 FEAT_B: 37ee8496 - Correlation: 0.9967764199761916\n",
      "241: FEAT_A: 30614231 FEAT_B: Cauldron_Filler__Assessment_ - Correlation: 0.99114888509877\n",
      "242: FEAT_A: 30614231 FEAT_B: Cauldron_Filler__Assessment__3110 - Correlation: 0.9914755860152119\n",
      "243: FEAT_A: 30614231 FEAT_B: Cauldron_Filler__Assessment__3010 - Correlation: 0.9915794451958683\n",
      "244: FEAT_A: 30614231 FEAT_B: Cauldron_Filler__Assessment__4020 - Correlation: 1.0\n",
      "245: FEAT_A: 30614231 FEAT_B: Cauldron_Filler__Assessment__4030 - Correlation: 0.9967764199761916\n",
      "246: FEAT_A: 6f4bd64e FEAT_B: Air_Show_4090 - Correlation: 1.0\n",
      "247: FEAT_A: f54238ee FEAT_B: Fireworks__Activity__4090 - Correlation: 1.0\n",
      "248: FEAT_A: f6947f54 FEAT_B: Bird_Measurer__Assessment__2030 - Correlation: 1.0\n",
      "249: FEAT_A: ad2fc29c FEAT_B: Bird_Measurer__Assessment__3020 - Correlation: 1.0\n",
      "250: FEAT_A: a8efe47b FEAT_B: Chest_Sorter__Assessment__4030 - Correlation: 1.0\n",
      "251: FEAT_A: 756e5507 FEAT_B: Chicken_Balancer__Activity__2000 - Correlation: 1.0\n",
      "252: FEAT_A: 7da34a02 FEAT_B: Mushroom_Sorter__Assessment__4070 - Correlation: 1.0\n",
      "253: FEAT_A: 611485c5 FEAT_B: Fireworks__Activity__4080 - Correlation: 1.0\n",
      "254: FEAT_A: 9d4e7b25 FEAT_B: Cart_Balancer__Assessment__4040 - Correlation: 0.9999999999999999\n",
      "255: FEAT_A: 0d1da71f FEAT_B: Chow_Time_3110 - Correlation: 1.0\n",
      "256: FEAT_A: 5c3d2b2f FEAT_B: Scrub_A_Dub - Correlation: 0.9911773750860648\n",
      "257: FEAT_A: 5c3d2b2f FEAT_B: Scrub_A_Dub_4020 - Correlation: 0.9999999999999998\n",
      "258: FEAT_A: c58186bf FEAT_B: Sandcastle_Builder__Activity__4035 - Correlation: 1.0\n",
      "259: FEAT_A: d3640339 FEAT_B: Dino_Dive_4090 - Correlation: 1.0\n",
      "260: FEAT_A: df4fe8b6 FEAT_B: 93b353f2 - Correlation: 0.9927227456002454\n",
      "261: FEAT_A: df4fe8b6 FEAT_B: ea296733 - Correlation: 0.9972490011446202\n",
      "262: FEAT_A: df4fe8b6 FEAT_B: Chest_Sorter__Assessment__4100 - Correlation: 0.9927227456002454\n",
      "263: FEAT_A: df4fe8b6 FEAT_B: Chest_Sorter__Assessment__3120 - Correlation: 0.9999999999999999\n",
      "264: FEAT_A: df4fe8b6 FEAT_B: Chest_Sorter__Assessment__3020 - Correlation: 0.9972490011446202\n",
      "265: FEAT_A: 56bcd38d FEAT_B: Chicken_Balancer__Activity_ - Correlation: 0.9905584109429233\n",
      "266: FEAT_A: 56bcd38d FEAT_B: Chicken_Balancer__Activity__4030 - Correlation: 1.0\n",
      "267: FEAT_A: 28f975ea FEAT_B: Air_Show_4020 - Correlation: 1.0\n",
      "268: FEAT_A: 15f99afc FEAT_B: 6cf7d25c - Correlation: 0.9994848288648872\n",
      "269: FEAT_A: 15f99afc FEAT_B: Pan_Balance_3110 - Correlation: 1.0\n",
      "270: FEAT_A: 15f99afc FEAT_B: Pan_Balance_3010 - Correlation: 0.9994848288648872\n",
      "271: FEAT_A: c2baf0bd FEAT_B: c7fe2a55 - Correlation: 0.9919981544366049\n",
      "272: FEAT_A: c2baf0bd FEAT_B: 36fa3ebe - Correlation: 0.9920334460178526\n",
      "273: FEAT_A: c2baf0bd FEAT_B: Happy_Camel_3021 - Correlation: 0.9919981544366049\n",
      "274: FEAT_A: c2baf0bd FEAT_B: Happy_Camel_2030 - Correlation: 0.9920334460178526\n",
      "275: FEAT_A: c2baf0bd FEAT_B: Happy_Camel_2020 - Correlation: 0.9999999999999999\n",
      "276: FEAT_A: ea321fb1 FEAT_B: 84b0e0c8 - Correlation: 0.9993007802002754\n",
      "277: FEAT_A: ea321fb1 FEAT_B: Chicken_Balancer__Activity__3010 - Correlation: 0.9999999999999999\n",
      "278: FEAT_A: ea321fb1 FEAT_B: Chicken_Balancer__Activity__3110 - Correlation: 0.9993007802002754\n",
      "279: FEAT_A: 5f0eb72c FEAT_B: fbaf3456 - Correlation: 0.9937068137497974\n",
      "280: FEAT_A: 5f0eb72c FEAT_B: Mushroom_Sorter__Assessment__4030 - Correlation: 0.9937068137497974\n",
      "281: FEAT_A: 5f0eb72c FEAT_B: Mushroom_Sorter__Assessment__4020 - Correlation: 1.0\n",
      "282: FEAT_A: 08ff79ad FEAT_B: Egg_Dropper__Activity__4090 - Correlation: 0.9999999999999999\n",
      "283: FEAT_A: 87d743c1 FEAT_B: Dino_Dive_4010 - Correlation: 1.0\n",
      "284: FEAT_A: 19967db1 FEAT_B: Chow_Time_4090 - Correlation: 1.0\n",
      "285: FEAT_A: 56cd3b43 FEAT_B: bbfe0445 - Correlation: 0.9996925958310808\n",
      "286: FEAT_A: 56cd3b43 FEAT_B: Flower_Waterer__Activity__3010 - Correlation: 1.0\n",
      "287: FEAT_A: 56cd3b43 FEAT_B: Flower_Waterer__Activity__3110 - Correlation: 0.9996925958310808\n",
      "288: FEAT_A: 14de4c5d FEAT_B: Air_Show_4100 - Correlation: 1.0\n",
      "289: FEAT_A: 47026d5f FEAT_B: cb6010f8 - Correlation: 0.9996901645044032\n",
      "290: FEAT_A: 47026d5f FEAT_B: 56817e2b - Correlation: 0.9993683874342136\n",
      "291: FEAT_A: 47026d5f FEAT_B: Chow_Time_2030 - Correlation: 0.9993683874342136\n",
      "292: FEAT_A: 47026d5f FEAT_B: Chow_Time_3121 - Correlation: 0.9996901645044032\n",
      "293: FEAT_A: 47026d5f FEAT_B: Chow_Time_3021 - Correlation: 1.0\n",
      "294: FEAT_A: 9554a50b FEAT_B: Cauldron_Filler__Assessment__4080 - Correlation: 0.9999999999999999\n",
      "295: FEAT_A: 99ea62f3 FEAT_B: Bubble_Bath_2083 - Correlation: 0.9999999999999999\n",
      "296: FEAT_A: 565a3990 FEAT_B: Bug_Measurer__Activity__4070 - Correlation: 1.0\n",
      "297: FEAT_A: f93fc684 FEAT_B: 7d093bf9 - Correlation: 0.9916711568592776\n",
      "298: FEAT_A: f93fc684 FEAT_B: Chow_Time_2000 - Correlation: 0.9916711568592776\n",
      "299: FEAT_A: f93fc684 FEAT_B: Chow_Time_4010 - Correlation: 1.0\n",
      "300: FEAT_A: 9ed8f6da FEAT_B: Dino_Drink_2075 - Correlation: 1.0\n",
      "301: FEAT_A: 26a5a3dd FEAT_B: All_Star_Sorting_4080 - Correlation: 1.0\n",
      "302: FEAT_A: 3afde5dd FEAT_B: b012cd7f - Correlation: 0.9999689265978794\n",
      "303: FEAT_A: 3afde5dd FEAT_B: e5c9df6f - Correlation: 0.99909107586748\n",
      "304: FEAT_A: 3afde5dd FEAT_B: Leaf_Leader_2030 - Correlation: 0.9999689265978794\n",
      "305: FEAT_A: 3afde5dd FEAT_B: Leaf_Leader_3021 - Correlation: 0.9999999999999999\n",
      "306: FEAT_A: 3afde5dd FEAT_B: Leaf_Leader_3121 - Correlation: 0.99909107586748\n",
      "307: FEAT_A: a1bbe385 FEAT_B: f28c589a - Correlation: 0.9999536800314802\n",
      "308: FEAT_A: a1bbe385 FEAT_B: Air_Show_3010 - Correlation: 0.9999536800314802\n",
      "309: FEAT_A: a1bbe385 FEAT_B: Air_Show_3110 - Correlation: 1.0\n",
      "310: FEAT_A: 3ccd3f02 FEAT_B: 3dcdda7f - Correlation: 0.9977338365647164\n",
      "311: FEAT_A: 3ccd3f02 FEAT_B: Chest_Sorter__Assessment__3010 - Correlation: 0.9977338365647164\n",
      "312: FEAT_A: 3ccd3f02 FEAT_B: Chest_Sorter__Assessment__3110 - Correlation: 1.0\n",
      "313: FEAT_A: 763fc34e FEAT_B: e57dd7af - Correlation: 0.9972722453150218\n",
      "314: FEAT_A: 763fc34e FEAT_B: Leaf_Leader_3020 - Correlation: 1.0\n",
      "315: FEAT_A: 763fc34e FEAT_B: Leaf_Leader_3120 - Correlation: 0.9972722453150218\n",
      "316: FEAT_A: 155f62a4 FEAT_B: 5b49460a - Correlation: 1.0\n",
      "317: FEAT_A: 155f62a4 FEAT_B: Chest_Sorter__Assessment__2000 - Correlation: 1.0\n",
      "318: FEAT_A: 155f62a4 FEAT_B: Chest_Sorter__Assessment__2020 - Correlation: 1.0\n",
      "319: FEAT_A: 6f8106d9 FEAT_B: Dino_Drink_4090 - Correlation: 1.0\n",
      "320: FEAT_A: e04fb33d FEAT_B: 7423acbc - Correlation: 0.9996289984253202\n",
      "321: FEAT_A: e04fb33d FEAT_B: Air_Show_3020 - Correlation: 0.9996289984253202\n",
      "322: FEAT_A: e04fb33d FEAT_B: Air_Show_3120 - Correlation: 0.9999999999999999\n",
      "323: FEAT_A: a8876db3 FEAT_B: Cart_Balancer__Assessment__3021 - Correlation: 1.0\n",
      "324: FEAT_A: 85d1b0de FEAT_B: Chicken_Balancer__Activity__4090 - Correlation: 1.0\n",
      "325: FEAT_A: daac11b0 FEAT_B: 1f19558b - Correlation: 0.9991356219818421\n",
      "326: FEAT_A: daac11b0 FEAT_B: ca11f653 - Correlation: 0.9995251391719705\n",
      "327: FEAT_A: daac11b0 FEAT_B: All_Star_Sorting_3021 - Correlation: 1.0\n",
      "328: FEAT_A: daac11b0 FEAT_B: All_Star_Sorting_3121 - Correlation: 0.9991356219818421\n",
      "329: FEAT_A: daac11b0 FEAT_B: All_Star_Sorting_2030 - Correlation: 0.9995251391719705\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "330: FEAT_A: 461eace6 FEAT_B: Egg_Dropper__Activity__4020 - Correlation: 1.0\n",
      "331: FEAT_A: 532a2afb FEAT_B: Cauldron_Filler__Assessment__2020 - Correlation: 1.0\n",
      "332: FEAT_A: d88ca108 FEAT_B: Air_Show_2070 - Correlation: 1.0\n",
      "333: FEAT_A: 7ab78247 FEAT_B: b80e5e84 - Correlation: 0.9998336639671014\n",
      "334: FEAT_A: 7ab78247 FEAT_B: Egg_Dropper__Activity__3010 - Correlation: 1.0\n",
      "335: FEAT_A: 7ab78247 FEAT_B: Egg_Dropper__Activity__3110 - Correlation: 0.9998336639671014\n",
      "336: FEAT_A: bfc77bd6 FEAT_B: Chest_Sorter__Assessment__4080 - Correlation: 1.0\n",
      "337: FEAT_A: 8b757ab8 FEAT_B: 44cb4907 - Correlation: 0.9998350601727668\n",
      "338: FEAT_A: 8b757ab8 FEAT_B: Crystals_Rule_3120 - Correlation: 1.0\n",
      "339: FEAT_A: 8b757ab8 FEAT_B: Crystals_Rule_3020 - Correlation: 0.9998350601727668\n",
      "340: FEAT_A: 5e812b27 FEAT_B: Sandcastle_Builder__Activity__4030 - Correlation: 1.0\n",
      "341: FEAT_A: 9e4c8c7b FEAT_B: 363d3849 - Correlation: 0.9992131086125489\n",
      "342: FEAT_A: 9e4c8c7b FEAT_B: All_Star_Sorting_3010 - Correlation: 0.9992131086125489\n",
      "343: FEAT_A: 9e4c8c7b FEAT_B: All_Star_Sorting_3110 - Correlation: 1.0\n",
      "344: FEAT_A: 0086365d FEAT_B: Pan_Balance_4010 - Correlation: 0.9999999999999999\n",
      "345: FEAT_A: 49ed92e9 FEAT_B: bd701df8 - Correlation: 0.9993109338825787\n",
      "346: FEAT_A: 49ed92e9 FEAT_B: Watering_Hole__Activity__3110 - Correlation: 0.9993109338825787\n",
      "347: FEAT_A: 49ed92e9 FEAT_B: Watering_Hole__Activity__3010 - Correlation: 1.0\n",
      "348: FEAT_A: a16a373e FEAT_B: Bird_Measurer__Assessment__4070 - Correlation: 1.0\n",
      "349: FEAT_A: a29c5338 FEAT_B: 7f0836bf - Correlation: 0.9986531994615934\n",
      "350: FEAT_A: a29c5338 FEAT_B: Dino_Drink_3010 - Correlation: 0.9999999999999999\n",
      "351: FEAT_A: a29c5338 FEAT_B: Dino_Drink_3110 - Correlation: 0.9986531994615934\n",
      "352: FEAT_A: abc5811c FEAT_B: Happy_Camel_4010 - Correlation: 0.9999999999999999\n",
      "353: FEAT_A: a44b10dc FEAT_B: Flower_Waterer__Activity__4070 - Correlation: 0.9999999999999999\n",
      "354: FEAT_A: 92687c59 FEAT_B: Scrub_A_Dub_4090 - Correlation: 0.9999999999999999\n",
      "355: FEAT_A: 1b54d27f FEAT_B: Watering_Hole__Activity__2010 - Correlation: 0.9999999999999999\n",
      "356: FEAT_A: 3393b68b FEAT_B: Bird_Measurer__Assessment__2010 - Correlation: 1.0\n",
      "357: FEAT_A: 5348fd84 FEAT_B: Cauldron_Filler__Assessment__4040 - Correlation: 1.0\n",
      "358: FEAT_A: f32856e4 FEAT_B: Leaf_Leader_2020 - Correlation: 1.0\n",
      "359: FEAT_A: d2659ab4 FEAT_B: Air_Show_2075 - Correlation: 1.0\n",
      "360: FEAT_A: 3a4be871 FEAT_B: Flower_Waterer__Activity__4080 - Correlation: 1.0\n",
      "361: FEAT_A: ec138c1c FEAT_B: Bird_Measurer__Assessment__2020 - Correlation: 1.0\n",
      "362: FEAT_A: 51102b85 FEAT_B: Bird_Measurer__Assessment__4030 - Correlation: 1.0\n",
      "363: FEAT_A: a1e4395d FEAT_B: Mushroom_Sorter__Assessment__3010 - Correlation: 1.0\n",
      "364: FEAT_A: 0db6d71d FEAT_B: Chest_Sorter__Assessment__4020 - Correlation: 0.9999999999999998\n",
      "365: FEAT_A: 13f56524 FEAT_B: Mushroom_Sorter__Assessment__4080 - Correlation: 1.0\n",
      "366: FEAT_A: 8d84fa81 FEAT_B: 99abe2bb - Correlation: 0.9910062251925293\n",
      "367: FEAT_A: 8d84fa81 FEAT_B: Bubble_Bath_4010 - Correlation: 1.0\n",
      "368: FEAT_A: 8d84fa81 FEAT_B: Bubble_Bath_2080 - Correlation: 0.9910062251925293\n",
      "369: FEAT_A: 77ead60d FEAT_B: 16dffff1 - Correlation: 0.9984675105119654\n",
      "370: FEAT_A: 77ead60d FEAT_B: 4d911100 - Correlation: 0.9998475946912412\n",
      "371: FEAT_A: 77ead60d FEAT_B: Dino_Drink_2030 - Correlation: 0.9984675105119654\n",
      "372: FEAT_A: 77ead60d FEAT_B: Dino_Drink_3021 - Correlation: 1.0\n",
      "373: FEAT_A: 77ead60d FEAT_B: Dino_Drink_3121 - Correlation: 0.9998475946912412\n",
      "374: FEAT_A: 29f54413 FEAT_B: Leaf_Leader_2060 - Correlation: 0.9999999999999998\n",
      "375: FEAT_A: e7561dd2 FEAT_B: Pan_Balance_4025 - Correlation: 0.9999999999999998\n",
      "376: FEAT_A: 8ac7cce4 FEAT_B: Leaf_Leader_2000 - Correlation: 1.0\n",
      "377: FEAT_A: 4bb2f698 FEAT_B: Chicken_Balancer__Activity__4070 - Correlation: 1.0\n",
      "378: FEAT_A: b120f2ac FEAT_B: d45ed6a1 - Correlation: 0.9979708133684774\n",
      "379: FEAT_A: b120f2ac FEAT_B: c277e121 - Correlation: 0.9999983835992123\n",
      "380: FEAT_A: b120f2ac FEAT_B: All_Star_Sorting_3120 - Correlation: 0.9979708133684774\n",
      "381: FEAT_A: b120f2ac FEAT_B: All_Star_Sorting_2025 - Correlation: 1.0\n",
      "382: FEAT_A: b120f2ac FEAT_B: All_Star_Sorting_3020 - Correlation: 0.9999983835992123\n",
      "383: FEAT_A: d185d3ea FEAT_B: Chow_Time_4035 - Correlation: 1.0\n",
      "384: FEAT_A: 5290eab1 FEAT_B: 04df9b66 - Correlation: 0.9998190483355675\n",
      "385: FEAT_A: 5290eab1 FEAT_B: Cauldron_Filler__Assessment__3120 - Correlation: 1.0\n",
      "386: FEAT_A: 5290eab1 FEAT_B: Cauldron_Filler__Assessment__3020 - Correlation: 0.9998190483355675\n",
      "387: FEAT_A: f50fc6c1 FEAT_B: Watering_Hole__Activity__4021 - Correlation: 1.0\n",
      "388: FEAT_A: 499edb7c FEAT_B: Chicken_Balancer__Activity__4020 - Correlation: 1.0\n",
      "389: FEAT_A: 4a09ace1 FEAT_B: Scrub_A_Dub_2083 - Correlation: 1.0\n",
      "390: FEAT_A: 47f43a44 FEAT_B: Flower_Waterer__Activity__4090 - Correlation: 0.9999999999999999\n",
      "391: FEAT_A: 01ca3a3c FEAT_B: Leaf_Leader_4080 - Correlation: 1.0\n",
      "392: FEAT_A: 37937459 FEAT_B: Sandcastle_Builder__Activity__4090 - Correlation: 0.9999999999999998\n",
      "393: FEAT_A: 37db1c2f FEAT_B: Happy_Camel_4045 - Correlation: 1.0\n",
      "394: FEAT_A: c7128948 FEAT_B: Mushroom_Sorter__Assessment__4040 - Correlation: 1.0\n",
      "395: FEAT_A: 0413e89d FEAT_B: Bubble_Bath_3010 - Correlation: 1.0\n",
      "396: FEAT_A: f3cd5473 FEAT_B: Pan_Balance_4070 - Correlation: 1.0\n",
      "397: FEAT_A: 46b50ba8 FEAT_B: Happy_Camel_4095 - Correlation: 1.0\n",
      "398: FEAT_A: cdd22e43 FEAT_B: Chicken_Balancer__Activity__4035 - Correlation: 1.0\n",
      "399: FEAT_A: 6f4adc4b FEAT_B: 55115cbd - Correlation: 0.9997831921803727\n",
      "400: FEAT_A: 6f4adc4b FEAT_B: Bubble_Bath_3121 - Correlation: 0.9997831921803727\n",
      "401: FEAT_A: 6f4adc4b FEAT_B: Bubble_Bath_3021 - Correlation: 1.0\n",
      "402: FEAT_A: 392e14df FEAT_B: Cauldron_Filler__Assessment__4100 - Correlation: 1.0\n",
      "403: FEAT_A: 25fa8af4 FEAT_B: Mushroom_Sorter__Assessment__4100 - Correlation: 0.9999999999999999\n",
      "404: FEAT_A: c7f7f0e1 FEAT_B: Bug_Measurer__Activity__2000 - Correlation: 1.0\n",
      "405: FEAT_A: 2a444e03 FEAT_B: 804ee27f - Correlation: 0.9947208630935935\n",
      "406: FEAT_A: 2a444e03 FEAT_B: Pan_Balance_4030 - Correlation: 1.0\n",
      "407: FEAT_A: 2a444e03 FEAT_B: Pan_Balance_4020 - Correlation: 0.9947208630935935\n",
      "408: FEAT_A: 47efca07 FEAT_B: Bottle_Filler__Activity__4090 - Correlation: 1.0\n",
      "409: FEAT_A: 0ce40006 FEAT_B: Happy_Camel_4080 - Correlation: 1.0\n",
      "410: FEAT_A: 9ce586dd FEAT_B: Chest_Sorter__Assessment__4035 - Correlation: 0.9999999999999999\n",
      "411: FEAT_A: 30df3273 FEAT_B: Sandcastle_Builder__Activity__4080 - Correlation: 0.9999999999999999\n",
      "412: FEAT_A: 3b2048ee FEAT_B: Leaf_Leader_4095 - Correlation: 1.0\n",
      "413: FEAT_A: ac92046e FEAT_B: d88e8f25 - Correlation: 0.9999763073610013\n",
      "414: FEAT_A: ac92046e FEAT_B: Scrub_A_Dub_3120 - Correlation: 1.0\n",
      "415: FEAT_A: ac92046e FEAT_B: Scrub_A_Dub_3020 - Correlation: 0.9999763073610013\n",
      "416: FEAT_A: 598f4598 FEAT_B: Flower_Waterer__Activity__4025 - Correlation: 0.9999999999999998\n",
      "417: FEAT_A: 1cf54632 FEAT_B: Bubble_Bath_2000 - Correlation: 1.0\n",
      "418: FEAT_A: 262136f4 FEAT_B: Leaf_Leader_4020 - Correlation: 1.0\n",
      "419: FEAT_A: 29a42aea FEAT_B: Bubble_Bath_4080 - Correlation: 1.0\n",
      "420: FEAT_A: 9e6b7fb5 FEAT_B: Chow_Time_4095 - Correlation: 1.0\n",
      "421: FEAT_A: 6aeafed4 FEAT_B: Bubble_Bath_4090 - Correlation: 0.9999999999999998\n",
      "422: FEAT_A: 7fd1ac25 FEAT_B: Egg_Dropper__Activity__4080 - Correlation: 0.9999999999999999\n",
      "423: FEAT_A: 63f13dd7 FEAT_B: Chow_Time_2020 - Correlation: 1.0\n",
      "424: FEAT_A: d9c005dd FEAT_B: Happy_Camel_2000 - Correlation: 1.0\n",
      "425: FEAT_A: 022b4259 FEAT_B: Bug_Measurer__Activity__4025 - Correlation: 0.9999999999999998\n",
      "426: FEAT_A: c189aaf2 FEAT_B: Happy_Camel_2083 - Correlation: 0.9999999999999999\n",
      "427: FEAT_A: 587b5989 FEAT_B: All_Star_Sorting_4070 - Correlation: 1.0\n",
      "428: FEAT_A: b7dc8128 FEAT_B: 4b5efe37 - Correlation: 0.9980151689807631\n",
      "429: FEAT_A: b7dc8128 FEAT_B: All_Star_Sorting_2000 - Correlation: 0.9999999999999999\n",
      "430: FEAT_A: b7dc8128 FEAT_B: All_Star_Sorting_4010 - Correlation: 0.9980151689807631\n",
      "431: FEAT_A: 5859dfb6 FEAT_B: 90ea0bac - Correlation: 0.9981052766128138\n",
      "432: FEAT_A: 5859dfb6 FEAT_B: Bubble_Bath_3020 - Correlation: 0.9981052766128138\n",
      "433: FEAT_A: 5859dfb6 FEAT_B: Bubble_Bath_3120 - Correlation: 1.0\n",
      "434: FEAT_A: 1bb5fbdb FEAT_B: b2dba42b - Correlation: 0.9999521739916128\n",
      "435: FEAT_A: 1bb5fbdb FEAT_B: Sandcastle_Builder__Activity__3110 - Correlation: 1.0\n",
      "436: FEAT_A: 1bb5fbdb FEAT_B: Sandcastle_Builder__Activity__3010 - Correlation: 0.9999521739916128\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "437: FEAT_A: 15ba1109 FEAT_B: Air_Show_2000 - Correlation: 1.0\n",
      "438: FEAT_A: e7e44842 FEAT_B: Watering_Hole__Activity__4090 - Correlation: 1.0\n",
      "439: FEAT_A: 46cd75b4 FEAT_B: Chicken_Balancer__Activity__4022 - Correlation: 1.0\n",
      "440: FEAT_A: d122731b FEAT_B: Cart_Balancer__Assessment__4100 - Correlation: 1.0\n",
      "441: FEAT_A: d2278a3b FEAT_B: Bottle_Filler__Activity__2000 - Correlation: 1.0\n",
      "442: FEAT_A: c0415e5c FEAT_B: Dino_Dive_4020 - Correlation: 1.0\n",
      "443: FEAT_A: 828e68f9 FEAT_B: 795e4a37 - Correlation: 0.993763490094275\n",
      "444: FEAT_A: 828e68f9 FEAT_B: Cart_Balancer__Assessment__3110 - Correlation: 0.9999999999999998\n",
      "445: FEAT_A: 828e68f9 FEAT_B: Cart_Balancer__Assessment__3010 - Correlation: 0.993763490094275\n",
      "446: FEAT_A: d3268efa FEAT_B: 28520915 - Correlation: 0.9989055302553063\n",
      "447: FEAT_A: d3268efa FEAT_B: b5053438 - Correlation: 0.9995763379622705\n",
      "448: FEAT_A: d3268efa FEAT_B: Cauldron_Filler__Assessment__3121 - Correlation: 0.9995763379622705\n",
      "449: FEAT_A: d3268efa FEAT_B: Cauldron_Filler__Assessment__3021 - Correlation: 0.9999999999999998\n",
      "450: FEAT_A: d3268efa FEAT_B: Cauldron_Filler__Assessment__2030 - Correlation: 0.9989055302553063\n",
      "451: FEAT_A: acf5c23f FEAT_B: Cart_Balancer__Assessment__4070 - Correlation: 1.0\n",
      "452: FEAT_A: 731c0cbe FEAT_B: Bird_Measurer__Assessment__4090 - Correlation: 1.0\n",
      "453: FEAT_A: 6f445b57 FEAT_B: Chow_Time_4080 - Correlation: 0.9999999999999998\n",
      "454: FEAT_A: 6077cc36 FEAT_B: Bird_Measurer__Assessment__4080 - Correlation: 1.0\n",
      "455: FEAT_A: e4d32835 FEAT_B: Pan_Balance_4080 - Correlation: 0.9999999999999998\n",
      "456: FEAT_A: 2b058fe3 FEAT_B: Cauldron_Filler__Assessment__2010 - Correlation: 1.0\n",
      "457: FEAT_A: 6bf9e3e1 FEAT_B: Happy_Camel_4040 - Correlation: 1.0\n",
      "458: FEAT_A: d51b1749 FEAT_B: Happy_Camel_2080 - Correlation: 0.9999999999999998\n",
      "459: FEAT_A: 119b5b02 FEAT_B: Dino_Dive_4080 - Correlation: 1.0\n",
      "460: FEAT_A: 070a5291 FEAT_B: Bird_Measurer__Assessment__4100 - Correlation: 0.9999999999999999\n",
      "461: FEAT_A: 562cec5f FEAT_B: Chest_Sorter__Assessment__4025 - Correlation: 1.0\n",
      "462: FEAT_A: 6088b756 FEAT_B: Dino_Dive_2070 - Correlation: 1.0\n",
      "463: FEAT_A: 7372e1a5 FEAT_B: Chow_Time_4070 - Correlation: 1.0\n",
      "464: FEAT_A: c1cac9a2 FEAT_B: Scrub_A_Dub_2081 - Correlation: 1.0\n",
      "465: FEAT_A: 9ee1c98c FEAT_B: Sandcastle_Builder__Activity__4021 - Correlation: 1.0\n",
      "466: FEAT_A: 84538528 FEAT_B: Sandcastle_Builder__Activity__4020 - Correlation: 0.9999999999999999\n",
      "467: FEAT_A: d38c2fd7 FEAT_B: Bird_Measurer__Assessment__4035 - Correlation: 1.0\n",
      "468: FEAT_A: 9c5ef70c FEAT_B: Pan_Balance_2000 - Correlation: 1.0\n",
      "469: FEAT_A: d3f1e122 FEAT_B: Bottle_Filler__Activity__4035 - Correlation: 0.9999999999999999\n",
      "470: FEAT_A: fd20ea40 FEAT_B: Leaf_Leader_4010 - Correlation: 1.0\n",
      "471: FEAT_A: ab4ec3a4 FEAT_B: Dino_Drink_4080 - Correlation: 1.0\n",
      "472: FEAT_A: cb1178ad FEAT_B: Chest_Sorter__Assessment__4090 - Correlation: 1.0\n",
      "473: FEAT_A: e64e2cfd FEAT_B: Watering_Hole__Activity__2000 - Correlation: 1.0\n",
      "474: FEAT_A: 77c76bc5 FEAT_B: Cauldron_Filler__Assessment__4090 - Correlation: 1.0\n",
      "475: FEAT_A: 8fee50e2 FEAT_B: Bird_Measurer__Assessment__4020 - Correlation: 1.0\n",
      "476: FEAT_A: 76babcde FEAT_B: Dino_Dive_4070 - Correlation: 1.0\n",
      "477: FEAT_A: 4a4c3d21 FEAT_B: Bird_Measurer__Assessment__4025 - Correlation: 0.9999999999999999\n",
      "478: FEAT_A: 65abac75 FEAT_B: Air_Show_4010 - Correlation: 0.9999999999999999\n",
      "479: FEAT_A: a7640a16 FEAT_B: Happy_Camel_4070 - Correlation: 1.0\n",
      "480: FEAT_A: 5be391b5 FEAT_B: Dino_Drink_4010 - Correlation: 1.0\n",
      "481: FEAT_A: bd612267 FEAT_B: Chest_Sorter__Assessment__4070 - Correlation: 1.0\n",
      "482: FEAT_A: a8a78786 FEAT_B: Happy_Camel_3121 - Correlation: 1.0\n",
      "483: FEAT_A: 4901243f FEAT_B: Fireworks__Activity__2000 - Correlation: 1.0\n",
      "484: FEAT_A: 48349b14 FEAT_B: cc5087a3 - Correlation: 0.9925680144288979\n",
      "485: FEAT_A: 48349b14 FEAT_B: Crystals_Rule_4010 - Correlation: 0.9925680144288979\n",
      "486: FEAT_A: 48349b14 FEAT_B: Crystals_Rule_2000 - Correlation: 1.0\n",
      "487: FEAT_A: 6c517a88 FEAT_B: Dino_Drink_4070 - Correlation: 1.0\n",
      "488: FEAT_A: 3d0b9317 FEAT_B: Chest_Sorter__Assessment__4040 - Correlation: 1.0\n",
      "489: FEAT_A: cf82af56 FEAT_B: Scrub_A_Dub_4070 - Correlation: 1.0\n",
      "490: FEAT_A: eb2c19cd FEAT_B: Mushroom_Sorter__Assessment__4090 - Correlation: 0.9999999999999999\n",
      "491: FEAT_A: d02b7a8e FEAT_B: All_Star_Sorting_4035 - Correlation: 1.0\n",
      "492: FEAT_A: 53c6e11a FEAT_B: Leaf_Leader_2075 - Correlation: 1.0\n",
      "493: FEAT_A: a5e9da97 FEAT_B: Pan_Balance_4100 - Correlation: 1.0\n",
      "494: FEAT_A: 1325467d FEAT_B: Sandcastle_Builder__Activity__4070 - Correlation: 1.0\n",
      "495: FEAT_A: 3ee399c3 FEAT_B: Cauldron_Filler__Assessment__4070 - Correlation: 0.9999999999999999\n",
      "496: FEAT_A: 2ec694de FEAT_B: Bug_Measurer__Activity__4080 - Correlation: 0.9999999999999998\n",
      "497: FEAT_A: dcb55a27 FEAT_B: Air_Show_4110 - Correlation: 0.9999999999999998\n",
      "498: FEAT_A: bb3e370b FEAT_B: Bottle_Filler__Activity_ - Correlation: 0.9950045049270008\n",
      "499: FEAT_A: bb3e370b FEAT_B: Bottle_Filler__Activity__4030 - Correlation: 1.0\n",
      "500: FEAT_A: de26c3a6 FEAT_B: Flower_Waterer__Activity__4020 - Correlation: 1.0\n",
      "501: FEAT_A: 4c2ec19f FEAT_B: Egg_Dropper__Activity__4025 - Correlation: 1.0\n",
      "502: FEAT_A: 857f21c0 FEAT_B: Bubble_Bath_4040 - Correlation: 1.0\n",
      "503: FEAT_A: 4ef8cdd3 FEAT_B: Chow_Time_4020 - Correlation: 1.0\n",
      "504: FEAT_A: 67aa2ada FEAT_B: Leaf_Leader_4090 - Correlation: 1.0\n",
      "505: FEAT_A: 06372577 FEAT_B: Air_Show_2060 - Correlation: 1.0\n",
      "506: FEAT_A: 91561152 FEAT_B: Cauldron_Filler__Assessment__4025 - Correlation: 1.0\n",
      "507: FEAT_A: 6043a2b4 FEAT_B: All_Star_Sorting_4090 - Correlation: 1.0\n",
      "508: FEAT_A: 86ba578b FEAT_B: Leaf_Leader_2070 - Correlation: 1.0\n",
      "509: FEAT_A: b1d5101d FEAT_B: All_Star_Sorting_4095 - Correlation: 0.9999999999999998\n",
      "510: FEAT_A: 884228c8 FEAT_B: Fireworks__Activity__4070 - Correlation: 1.0\n",
      "511: FEAT_A: 792530f8 FEAT_B: Dino_Drink - Correlation: 0.9905196877266726\n",
      "512: FEAT_A: 792530f8 FEAT_B: Dino_Drink_4030 - Correlation: 1.0\n",
      "513: FEAT_A: Crystal_Caves___Level_3 FEAT_B: Crystal_Caves___Level_3_2000 - Correlation: 1.0\n",
      "514: FEAT_A: Slop_Problem FEAT_B: Slop_Problem_2000 - Correlation: 1.0\n",
      "515: FEAT_A: Treasure_Map FEAT_B: Treasure_Map_2000 - Correlation: 1.0\n",
      "516: FEAT_A: Balancing_Act FEAT_B: Balancing_Act_2000 - Correlation: 1.0\n",
      "517: FEAT_A: Lifting_Heavy_Things FEAT_B: Lifting_Heavy_Things_2000 - Correlation: 1.0\n",
      "518: FEAT_A: Crystal_Caves___Level_1 FEAT_B: Crystal_Caves___Level_1_2000 - Correlation: 1.0\n",
      "519: FEAT_A: Pirate_s_Tale FEAT_B: Pirate_s_Tale_2000 - Correlation: 1.0\n",
      "520: FEAT_A: Tree_Top_City___Level_2 FEAT_B: Tree_Top_City___Level_2_2000 - Correlation: 1.0\n",
      "521: FEAT_A: Tree_Top_City___Level_3 FEAT_B: Tree_Top_City___Level_3_2000 - Correlation: 1.0\n",
      "522: FEAT_A: Heavy__Heavier__Heaviest FEAT_B: Heavy__Heavier__Heaviest_2000 - Correlation: 1.0\n",
      "523: FEAT_A: Ordering_Spheres FEAT_B: Ordering_Spheres_2000 - Correlation: 1.0\n",
      "524: FEAT_A: 12_Monkeys FEAT_B: 12_Monkeys_2000 - Correlation: 1.0\n",
      "525: FEAT_A: Magma_Peak___Level_2 FEAT_B: Magma_Peak___Level_2_2000 - Correlation: 0.9999999999999999\n",
      "526: FEAT_A: Honey_Cake FEAT_B: Honey_Cake_2000 - Correlation: 1.0\n",
      "527: FEAT_A: Rulers FEAT_B: Rulers_2000 - Correlation: 0.9999999999999998\n",
      "528: FEAT_A: Magma_Peak___Level_1 FEAT_B: Magma_Peak___Level_1_2000 - Correlation: 0.9999999999999999\n",
      "529: FEAT_A: Tree_Top_City___Level_1 FEAT_B: Tree_Top_City___Level_1_2000 - Correlation: 1.0\n",
      "530: FEAT_A: Costume_Box FEAT_B: Costume_Box_2000 - Correlation: 1.0\n",
      "531: FEAT_A: Crystal_Caves___Level_2 FEAT_B: Crystal_Caves___Level_2_2000 - Correlation: 1.0\n",
      "532: FEAT_A: var_event_id FEAT_B: var_title_event_code - Correlation: 0.9995254853903907\n",
      "533: FEAT_A: accumulated_accuracy FEAT_B: accumulated_accuracy_group - Correlation: 0.994549093468193\n"
     ]
    }
   ],
   "source": [
    "# Selecting high-correlated features\n",
    "counter = 0\n",
    "to_remove = []\n",
    "for feat_a in features:\n",
    "    for feat_b in features:\n",
    "        if feat_a != feat_b and feat_a not in to_remove and feat_b not in to_remove:\n",
    "            c = np.corrcoef(new_train[feat_a], new_train[feat_b])[0][1]\n",
    "            if c > 0.99:\n",
    "                counter += 1\n",
    "                to_remove.append(feat_b)\n",
    "                print('{}: FEAT_A: {} FEAT_B: {} - Correlation: {}'.format(counter, feat_a, feat_b, c))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total Number of Removals: 533\n",
      "------------------------------\n"
     ]
    }
   ],
   "source": [
    "print('Total Number of Removals:', counter)\n",
    "print('-'* 30)\n",
    "# print('Removed Features:', to_remove)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of features after removal:  390\n"
     ]
    }
   ],
   "source": [
    "features = [x for x in features if x not in to_remove]\n",
    "print(\"Number of features after removal: \", len(features))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Regression Modelling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def eval_qwk_lgb_regr(y_true, y_pred):\n",
    "    \"\"\"\n",
    "    Fast qwk eval function for lgb.\n",
    "    \"\"\"\n",
    "    dist = Counter(new_train['accuracy_group'])\n",
    "    for k in dist:\n",
    "        dist[k] /= len(new_train)\n",
    "    new_train['accuracy_group'].hist()\n",
    "    \n",
    "    acum = 0\n",
    "    bound = {}\n",
    "    for i in range(3):\n",
    "        acum += dist[i]\n",
    "        bound[i] = np.percentile(y_pred, acum * 100)\n",
    "\n",
    "    def classify(x):\n",
    "        if x <= bound[0]:\n",
    "            return 0\n",
    "        elif x <= bound[1]:\n",
    "            return 1\n",
    "        elif x <= bound[2]:\n",
    "            return 2\n",
    "        else:\n",
    "            return 3\n",
    "\n",
    "    y_pred = np.array(list(map(classify, y_pred))).reshape(y_true.shape)\n",
    "\n",
    "    return 'cappa', cohen_kappa_score(y_true, y_pred, weights='quadratic'), True\n",
    "\n",
    "\n",
    "def qwk(act,pred,n=4,hist_range=(0,3)):\n",
    "    \"\"\"\n",
    "    QWK for normal use \n",
    "    \"\"\"\n",
    "    if pred.dtype != act.dtype:\n",
    "        pred = np.round(pred).astype('int')\n",
    "    O = confusion_matrix(act,pred)\n",
    "    O = np.divide(O,np.sum(O))\n",
    "    \n",
    "    W = np.zeros((n,n))\n",
    "    for i in range(n):\n",
    "        for j in range(n):\n",
    "            W[i][j] = ((i-j)**2)/((n-1)**2)\n",
    "            \n",
    "    act_hist = np.histogram(act,bins=n,range=hist_range)[0]\n",
    "    prd_hist = np.histogram(pred,bins=n,range=hist_range)[0]\n",
    "    \n",
    "    E = np.outer(act_hist,prd_hist)\n",
    "    E = np.divide(E,np.sum(E))\n",
    "    \n",
    "    num = np.sum(np.multiply(W,O))\n",
    "    den = np.sum(np.multiply(W,E))\n",
    "    print('Done scoring!')\n",
    "    return 1-np.divide(num,den)\n",
    "\n",
    "qwk_scorer = make_scorer(qwk, greater_is_better=True) # second param is used to differentiate loss and score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Base_Model(object):\n",
    "    \n",
    "    def __init__(self, train_df, test_df, features, categoricals=[], n_splits=5, verbose=True):\n",
    "        self.train_df = train_df\n",
    "        self.test_df = test_df\n",
    "        self.features = features\n",
    "        self.n_splits = n_splits\n",
    "        self.categoricals = categoricals\n",
    "        self.target = 'accuracy_group'\n",
    "        self.cv = self.get_group_cv() # could be either group/stratified k-fold\n",
    "        self.verbose = verbose\n",
    "        self.params = self.get_params()\n",
    "        self.y_pred, self.score, self.model = self.fit()\n",
    "        \n",
    "    def train_model(self, train_set, val_set):\n",
    "        raise NotImplementedError\n",
    "        \n",
    "    def get_stratified_cv(self):\n",
    "        cv = StratifiedKFold(n_splits=self.n_splits, shuffle=True, random_state=42)\n",
    "        return cv.split(self.train_df, self.train_df[self.target])\n",
    "    \n",
    "    def get_group_cv(self):\n",
    "        groups = new_train['installation_id'] # group by installation_id\n",
    "        cv = GroupKFold(n_splits=self.n_splits)\n",
    "        return cv.split(self.train_df, self.train_df[self.target], groups)\n",
    "    \n",
    "    def get_params(self):\n",
    "        raise NotImplementedError\n",
    "        \n",
    "    def convert_dataset(self, x_train, y_train, x_val, y_val):\n",
    "        raise NotImplementedError\n",
    "        \n",
    "    def convert_x(self, x):\n",
    "        return x\n",
    "        \n",
    "    def fit(self):\n",
    "        oof_pred = np.zeros((len(self.train_df), ))\n",
    "        y_pred = np.zeros((len(self.test_df), ))\n",
    "        \n",
    "        for fold, (train_idx, val_idx) in enumerate(self.cv):\n",
    "            x_train, x_val = self.train_df[self.features].iloc[train_idx], self.train_df[self.features].iloc[val_idx]\n",
    "            y_train, y_val = self.train_df[self.target][train_idx], self.train_df[self.target][val_idx]\n",
    "            train_set, val_set = self.convert_dataset(x_train, y_train, x_val, y_val)\n",
    "            model = self.train_model(train_set, val_set)\n",
    "            conv_x_val = self.convert_x(x_val)\n",
    "            oof_pred[val_idx] = model.predict(conv_x_val).reshape(oof_pred[val_idx].shape)\n",
    "            x_test = self.convert_x(self.test_df[self.features])\n",
    "            y_pred += model.predict(x_test).reshape(y_pred.shape) / self.n_splits     \n",
    "\n",
    "            print('-'*30)\n",
    "            print('Partial score of fold {} is: {}'.format(fold, eval_qwk_lgb_regr(y_val, oof_pred[val_idx])[1]))\n",
    "            print('-'*30)\n",
    "        _, loss_score, _ = eval_qwk_lgb_regr(self.train_df[self.target], oof_pred)\n",
    "        if self.verbose:\n",
    "            print('Our oof cohen kappa score is: ', loss_score)\n",
    "        return y_pred, loss_score, model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Catb_Model(Base_Model):\n",
    "    \n",
    "    def train_model(self, train_set, val_set):\n",
    "        verbosity = 500 if self.verbose else 0\n",
    "        clf = CatBoostRegressor(**self.params)\n",
    "        clf.fit(train_set['X'], \n",
    "                train_set['y'], \n",
    "                eval_set=(val_set['X'], val_set['y']),\n",
    "                verbose=verbosity, \n",
    "                cat_features=self.categoricals,\n",
    "                plot=False)\n",
    "        return clf\n",
    "        \n",
    "    def convert_dataset(self, x_train, y_train, x_val, y_val):\n",
    "        train_set = {'X': x_train, 'y': y_train}\n",
    "        val_set = {'X': x_val, 'y': y_val}\n",
    "        return train_set, val_set\n",
    "        \n",
    "    def get_params(self):\n",
    "        params = {'loss_function': 'RMSE',\n",
    "                  'task_type': \"CPU\",\n",
    "                  'iterations': 5000,\n",
    "                  'od_type': \"Iter\",\n",
    "                  'depth': 10,\n",
    "                  # 'learning_rate': 0.01,\n",
    "                  'colsample_bylevel': 0.5, \n",
    "                  'early_stopping_rounds': 100,\n",
    "                  'l2_leaf_reg': 18,\n",
    "                  'random_seed': 42,\n",
    "                  'use_best_model': True,\n",
    "                }\n",
    "        return params"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Lgb_Model(Base_Model):\n",
    "    \n",
    "    def train_model(self, train_set, val_set):\n",
    "        verbosity = 500 if self.verbose else 0\n",
    "        return lgb.train(self.params, train_set, valid_sets=[train_set, val_set], verbose_eval=verbosity)\n",
    "        \n",
    "    def convert_dataset(self, x_train, y_train, x_val, y_val):\n",
    "        train_set = lgb.Dataset(x_train, y_train, categorical_feature=self.categoricals)\n",
    "        val_set = lgb.Dataset(x_val, y_val, categorical_feature=self.categoricals)\n",
    "        return train_set, val_set\n",
    "        \n",
    "    def get_params(self):\n",
    "        params = {'n_estimators':5000,\n",
    "                    'boosting_type': 'gbdt',\n",
    "                    'objective': 'regression',\n",
    "                    'metric': 'rmse',\n",
    "                    'subsample': 0.75,\n",
    "                    'subsample_freq': 1,\n",
    "                    'learning_rate': 0.02,\n",
    "                    'feature_fraction': 0.9,\n",
    "                    'max_depth': 16,\n",
    "                    'lambda_l1': 1,  \n",
    "                    'lambda_l2': 1,\n",
    "                    'early_stopping_rounds': 100\n",
    "                    }\n",
    "        return params"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Xgb_Model(Base_Model):\n",
    "    \n",
    "    def train_model(self, train_set, val_set):\n",
    "        verbosity = 100 if self.verbose else 0\n",
    "        return xgb.train(self.params, train_set, \n",
    "                         num_boost_round=5000, evals=[(train_set, 'train'), (val_set, 'val')], \n",
    "                         verbose_eval=verbosity, early_stopping_rounds=100)\n",
    "        \n",
    "    def convert_dataset(self, x_train, y_train, x_val, y_val):\n",
    "        train_set = xgb.DMatrix(x_train, y_train)\n",
    "        val_set = xgb.DMatrix(x_val, y_val)\n",
    "        return train_set, val_set\n",
    "    \n",
    "    def convert_x(self, x):\n",
    "        return xgb.DMatrix(x)\n",
    "        \n",
    "    def get_params(self):\n",
    "        params = {'colsample_bytree': 0.8,                 \n",
    "            'learning_rate': 0.02,\n",
    "            'max_depth': 10,\n",
    "            'subsample': 1,\n",
    "            'objective':'reg:squarederror',\n",
    "            #'eval_metric':'rmse',\n",
    "            'min_child_weight':3,\n",
    "            'gamma':0.25,\n",
    "            'n_estimators':5000}\n",
    "\n",
    "        return params"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Nn_Model(Base_Model):\n",
    "    \n",
    "    def __init__(self, train_df, test_df, features, categoricals=[], n_splits=5, verbose=True):\n",
    "        features = features.copy()\n",
    "        if len(categoricals) > 0:\n",
    "            for cat in categoricals:\n",
    "                enc = OneHotEncoder()\n",
    "                train_cats = enc.fit_transform(train_df[[cat]])\n",
    "                test_cats = enc.transform(test_df[[cat]])\n",
    "                cat_cols = enc.get_feature_names()\n",
    "                features.extend(cat_cols)\n",
    "                train_cats = pd.DataFrame(train_cats.toarray(), columns=cat_cols)\n",
    "                test_cats = pd.DataFrame(test_cats.toarray(), columns=cat_cols)\n",
    "                train_df = pd.concat([train_df, train_cats], axis=1)\n",
    "                test_df = pd.concat([test_df, test_cats], axis=1)\n",
    "        scalar = MinMaxScaler()\n",
    "        train_df[features] = scalar.fit_transform(train_df[features])\n",
    "        test_df[features] = scalar.transform(test_df[features])\n",
    "        print(train_df[features].shape)\n",
    "        super().__init__(train_df, test_df, features, categoricals, n_splits, verbose)\n",
    "        \n",
    "    def train_model(self, train_set, val_set):\n",
    "        verbosity = 100 if self.verbose else 0\n",
    "        model = tf.keras.models.Sequential([\n",
    "            tf.keras.layers.Input(shape=(train_set['X'].shape[1],)),\n",
    "            tf.keras.layers.Dense(200, activation='relu'),\n",
    "            tf.keras.layers.LayerNormalization(),\n",
    "            tf.keras.layers.Dropout(0.3),\n",
    "            tf.keras.layers.Dense(100, activation='relu'),\n",
    "            tf.keras.layers.LayerNormalization(),\n",
    "            tf.keras.layers.Dropout(0.3),\n",
    "            tf.keras.layers.Dense(50, activation='relu'),\n",
    "            tf.keras.layers.LayerNormalization(),\n",
    "            tf.keras.layers.Dropout(0.3),\n",
    "            tf.keras.layers.Dense(25, activation='relu'),\n",
    "            tf.keras.layers.LayerNormalization(),\n",
    "            tf.keras.layers.Dropout(0.3),\n",
    "            tf.keras.layers.Dense(1, activation='relu')\n",
    "        ])\n",
    "        model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=4e-4), loss='mse')\n",
    "        print(model.summary())\n",
    "        save_best = tf.keras.callbacks.ModelCheckpoint('nn_model.w8', save_weights_only=True, save_best_only=True, verbose=1)\n",
    "        early_stop = tf.keras.callbacks.EarlyStopping(patience=20)\n",
    "        model.fit(train_set['X'], \n",
    "                train_set['y'], \n",
    "                validation_data=(val_set['X'], val_set['y']),\n",
    "                epochs=100,\n",
    "                 callbacks=[save_best, early_stop])\n",
    "        model.load_weights('nn_model.w8')\n",
    "        return model\n",
    "        \n",
    "    def convert_dataset(self, x_train, y_train, x_val, y_val):\n",
    "        train_set = {'X': x_train, 'y': y_train}\n",
    "        val_set = {'X': x_val, 'y': y_val}\n",
    "        return train_set, val_set\n",
    "        \n",
    "    def get_params(self):\n",
    "        return None"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training until validation scores don't improve for 100 rounds\n",
      "[500]\ttraining's rmse: 0.822805\tvalid_1's rmse: 0.966229\n",
      "Early stopping, best iteration is:\n",
      "[520]\ttraining's rmse: 0.818127\tvalid_1's rmse: 0.96546\n",
      "------------------------------\n",
      "Partial score of fold 0 is: 0.6194555665178476\n",
      "------------------------------\n",
      "Training until validation scores don't improve for 100 rounds\n",
      "[500]\ttraining's rmse: 0.818931\tvalid_1's rmse: 1.00219\n",
      "Early stopping, best iteration is:\n",
      "[401]\ttraining's rmse: 0.844216\tvalid_1's rmse: 1.00184\n",
      "------------------------------\n",
      "Partial score of fold 1 is: 0.5912009905548221\n",
      "------------------------------\n",
      "Training until validation scores don't improve for 100 rounds\n",
      "Early stopping, best iteration is:\n",
      "[359]\ttraining's rmse: 0.858472\tvalid_1's rmse: 0.979958\n",
      "------------------------------\n",
      "Partial score of fold 2 is: 0.5699527843863459\n",
      "------------------------------\n",
      "Training until validation scores don't improve for 100 rounds\n",
      "Early stopping, best iteration is:\n",
      "[383]\ttraining's rmse: 0.84503\tvalid_1's rmse: 0.999975\n",
      "------------------------------\n",
      "Partial score of fold 3 is: 0.5768284288543732\n",
      "------------------------------\n",
      "Training until validation scores don't improve for 100 rounds\n",
      "Early stopping, best iteration is:\n",
      "[376]\ttraining's rmse: 0.841492\tvalid_1's rmse: 1.01897\n",
      "------------------------------\n",
      "Partial score of fold 4 is: 0.5553262333676721\n",
      "------------------------------\n",
      "Our oof cohen kappa score is:  0.5838631095389941\n",
      "[0]\ttrain-rmse:1.84281\tval-rmse:1.84014\n",
      "Multiple eval metrics have been passed: 'val-rmse' will be used for early stopping.\n",
      "\n",
      "Will train until val-rmse hasn't improved in 100 rounds.\n",
      "[100]\ttrain-rmse:0.724737\tval-rmse:1.02267\n",
      "[200]\ttrain-rmse:0.563608\tval-rmse:0.993555\n",
      "[300]\ttrain-rmse:0.512832\tval-rmse:0.995915\n",
      "Stopping. Best iteration:\n",
      "[206]\ttrain-rmse:0.559583\tval-rmse:0.993481\n",
      "\n",
      "------------------------------\n",
      "Partial score of fold 0 is: 0.5963329788325569\n",
      "------------------------------\n",
      "[0]\ttrain-rmse:1.83887\tval-rmse:1.85594\n",
      "Multiple eval metrics have been passed: 'val-rmse' will be used for early stopping.\n",
      "\n",
      "Will train until val-rmse hasn't improved in 100 rounds.\n",
      "[100]\ttrain-rmse:0.725018\tval-rmse:1.05892\n",
      "[200]\ttrain-rmse:0.569037\tval-rmse:1.0316\n",
      "[300]\ttrain-rmse:0.517004\tval-rmse:1.03055\n",
      "[400]\ttrain-rmse:0.472826\tval-rmse:1.03179\n",
      "Stopping. Best iteration:\n",
      "[306]\ttrain-rmse:0.514209\tval-rmse:1.03036\n",
      "\n",
      "------------------------------\n",
      "Partial score of fold 1 is: 0.5489634545235861\n",
      "------------------------------\n",
      "[0]\ttrain-rmse:1.8328\tval-rmse:1.87991\n",
      "Multiple eval metrics have been passed: 'val-rmse' will be used for early stopping.\n",
      "\n",
      "Will train until val-rmse hasn't improved in 100 rounds.\n",
      "[100]\ttrain-rmse:0.740616\tval-rmse:1.04008\n",
      "[200]\ttrain-rmse:0.575765\tval-rmse:1.00599\n",
      "[300]\ttrain-rmse:0.515145\tval-rmse:1.00581\n",
      "Stopping. Best iteration:\n",
      "[239]\ttrain-rmse:0.548454\tval-rmse:1.0044\n",
      "\n",
      "------------------------------\n",
      "Partial score of fold 2 is: 0.543939278692144\n",
      "------------------------------\n",
      "[0]\ttrain-rmse:1.85193\tval-rmse:1.8029\n",
      "Multiple eval metrics have been passed: 'val-rmse' will be used for early stopping.\n",
      "\n",
      "Will train until val-rmse hasn't improved in 100 rounds.\n",
      "[100]\ttrain-rmse:0.728282\tval-rmse:1.04442\n",
      "[200]\ttrain-rmse:0.568151\tval-rmse:1.02276\n",
      "[300]\ttrain-rmse:0.510476\tval-rmse:1.02197\n",
      "Stopping. Best iteration:\n",
      "[242]\ttrain-rmse:0.540059\tval-rmse:1.02165\n",
      "\n",
      "------------------------------\n",
      "Partial score of fold 3 is: 0.5566238476563068\n",
      "------------------------------\n",
      "[0]\ttrain-rmse:1.8422\tval-rmse:1.8423\n",
      "Multiple eval metrics have been passed: 'val-rmse' will be used for early stopping.\n",
      "\n",
      "Will train until val-rmse hasn't improved in 100 rounds.\n",
      "[100]\ttrain-rmse:0.725255\tval-rmse:1.05647\n",
      "[200]\ttrain-rmse:0.556169\tval-rmse:1.0364\n",
      "[300]\ttrain-rmse:0.489055\tval-rmse:1.037\n",
      "Stopping. Best iteration:\n",
      "[231]\ttrain-rmse:0.529402\tval-rmse:1.03599\n",
      "\n",
      "------------------------------\n",
      "Partial score of fold 4 is: 0.5374714344665545\n",
      "------------------------------\n",
      "Our oof cohen kappa score is:  0.5575638305631696\n",
      "0:\tlearn: 1.2463872\ttest: 1.2551490\tbest: 1.2551490 (0)\ttotal: 283ms\tremaining: 23m 36s\n",
      "500:\tlearn: 0.8684071\ttest: 0.9829492\tbest: 0.9829131 (499)\ttotal: 2m 3s\tremaining: 18m 25s\n",
      "1000:\tlearn: 0.7825484\ttest: 0.9783378\tbest: 0.9782380 (990)\ttotal: 4m\tremaining: 16m\n",
      "Stopped by overfitting detector  (100 iterations wait)\n",
      "\n",
      "bestTest = 0.976502356\n",
      "bestIteration = 1314\n",
      "\n",
      "Shrink model to first 1315 iterations.\n",
      "------------------------------\n",
      "Partial score of fold 0 is: 0.616965441690201\n",
      "------------------------------\n",
      "0:\tlearn: 1.2445694\ttest: 1.2578627\tbest: 1.2578627 (0)\ttotal: 250ms\tremaining: 20m 47s\n",
      "500:\tlearn: 0.8779618\ttest: 1.0136470\tbest: 1.0136470 (500)\ttotal: 1m 45s\tremaining: 15m 47s\n",
      "1000:\tlearn: 0.7933132\ttest: 1.0089904\tbest: 1.0089408 (995)\ttotal: 3m 27s\tremaining: 13m 50s\n",
      "Stopped by overfitting detector  (100 iterations wait)\n",
      "\n",
      "bestTest = 1.008850406\n",
      "bestIteration = 1079\n",
      "\n",
      "Shrink model to first 1080 iterations.\n",
      "------------------------------\n",
      "Partial score of fold 1 is: 0.5757612189803787\n",
      "------------------------------\n",
      "0:\tlearn: 1.2544632\ttest: 1.2198003\tbest: 1.2198003 (0)\ttotal: 242ms\tremaining: 20m 9s\n",
      "500:\tlearn: 0.8751046\ttest: 0.9895341\tbest: 0.9894777 (498)\ttotal: 1m 49s\tremaining: 16m 27s\n",
      "Stopped by overfitting detector  (100 iterations wait)\n",
      "\n",
      "bestTest = 0.9879356893\n",
      "bestIteration = 736\n",
      "\n",
      "Shrink model to first 737 iterations.\n",
      "------------------------------\n",
      "Partial score of fold 2 is: 0.5662889103449089\n",
      "------------------------------\n",
      "0:\tlearn: 1.2445830\ttest: 1.2584683\tbest: 1.2584683 (0)\ttotal: 231ms\tremaining: 19m 13s\n",
      "500:\tlearn: 0.8675197\ttest: 1.0097848\tbest: 1.0097218 (499)\ttotal: 1m 48s\tremaining: 16m 13s\n",
      "Stopped by overfitting detector  (100 iterations wait)\n",
      "\n",
      "bestTest = 1.006934477\n",
      "bestIteration = 805\n",
      "\n",
      "Shrink model to first 806 iterations.\n",
      "------------------------------\n",
      "Partial score of fold 3 is: 0.5764733446746833\n",
      "------------------------------\n",
      "0:\tlearn: 1.2468430\ttest: 1.2515969\tbest: 1.2515969 (0)\ttotal: 246ms\tremaining: 20m 29s\n",
      "500:\tlearn: 0.8556919\ttest: 1.0288195\tbest: 1.0286472 (492)\ttotal: 1m 49s\tremaining: 16m 23s\n",
      "1000:\tlearn: 0.7643077\ttest: 1.0271647\tbest: 1.0262917 (909)\ttotal: 3m 33s\tremaining: 14m 14s\n",
      "Stopped by overfitting detector  (100 iterations wait)\n",
      "\n",
      "bestTest = 1.026291748\n",
      "bestIteration = 909\n",
      "\n",
      "Shrink model to first 910 iterations.\n",
      "------------------------------\n",
      "Partial score of fold 4 is: 0.5524694655434932\n",
      "------------------------------\n",
      "Our oof cohen kappa score is:  0.5763490298316156\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAD4CAYAAAAAczaOAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAARA0lEQVR4nO3dfYxcV3nH8e+DnThAAg6Ebi3brYOwtg2otGFlTJHQglHipBWO1ESyVYFBQZbaAKGqVF6k1i4QCSRECm0BuXUqByE7qUGNm4ZSN8kI9Y8Y4hBegnGzDWriZksAO4blJdTo6R9zNmyc2d27u7MzO3u+H2nle88998555iS/uXvnzmxkJpKkOjyn3wOQJPWOoS9JFTH0Jakihr4kVcTQl6SKrOz3AGZyySWX5IYNG+a9/49//GOe//znd29AfbJc6gBrWYqWSx1gLZOOHTv2/cx8SadtSzr0N2zYwP333z/v/VutFqOjo90bUJ8slzrAWpai5VIHWMukiPjv6bZ5eUeSKmLoS1JFDH1JqoihL0kVMfQlqSKGviRVxNCXpIoY+pJUEUNfkiqypD+RK0n9tGfPnr499mJ9stgzfUmqiKEvSRUx9CWpIoa+JFXE0Jekihj6klQRQ1+SKmLoS1JFDH1JqoihL0kVMfQlqSKGviRVxNCXpIoY+pJUEUNfkirSKPQj4k8i4qGI+GZEHIiICyLi0og4GhEPR8RtEXF+6buqrI+V7RumHOd9pf1ERFy5OCVJkqYza+hHxFrgXcBIZr4CWAFsBz4C3JyZG4HTwPVll+uB05n5MuDm0o+IuKzs93JgK/DJiFjR3XIkSTNpenlnJfDciFgJPA8YB94AHCrb9wPXlOVtZZ2yfUtERGk/mJlPZeZ3gDFg08JLkCQ1NeufS8zM/4mIjwKPAj8F/g04BjyZmWdLt5PA2rK8Fnis7Hs2Is4ALy7t90059NR9nhYRu4BdAENDQ7RarblXVUxMTCxo/6ViudQB1rIULZc6oPu1DA8Pd+1Yc7VY8zJr6EfExbTP0i8FngT+EbiqQ9ec3GWabdO1P7Mhcy+wF2BkZCQX8nciW63Wov2dyV5aLnWAtSxFy6UO6H4t/f4buYsxL00u77wR+E5mfi8z/w/4PPC7wOpyuQdgHfB4WT4JrAco218InJra3mEfSVIPNAn9R4HNEfG8cm1+C/At4F7g2tJnJ3BHWT5c1inb78nMLO3by909lwIbgS93pwxJUhNNrukfjYhDwAPAWeCrtC+//AtwMCI+VNr2lV32AZ+JiDHaZ/jby3Eeiojbab9gnAVuyMxfdLkeSdIMZg19gMzcDew+p/kROtx9k5k/A66b5jg3ATfNcYySpC7xE7mSVBFDX5IqYuhLUkUMfUmqiKEvSRUx9CWpIoa+JFXE0Jekihj6klQRQ1+SKmLoS1JFDH1JqoihL0kVMfQlqSKGviRVxNCXpIoY+pJUEUNfkipi6EtSRQx9SaqIoS9JFTH0Jakihr4kVcTQl6SKGPqSVBFDX5IqYuhLUkUMfUmqiKEvSRUx9CWpIoa+JFXE0Jekihj6klQRQ1+SKmLoS1JFDH1JqoihL0kVMfQlqSKNQj8iVkfEoYj4dkQcj4jXRMSLIuJIRDxc/r249I2I+EREjEXE1yPi8inH2Vn6PxwROxerKElSZ03P9D8O/Gtm/gbwSuA48F7g7szcCNxd1gGuAjaWn13ApwAi4kXAbuDVwCZg9+QLhSSpN2YN/Yh4AfA6YB9AZv48M58EtgH7S7f9wDVleRtwa7bdB6yOiDXAlcCRzDyVmaeBI8DWrlYjSZrRygZ9Xgp8D/iHiHglcAy4ERjKzHGAzByPiF8p/dcCj03Z/2Rpm679GSJiF+3fEBgaGqLVas2lnmeYmJhY0P5LxXKpA6xlKVoudUD3axkeHu7aseZqsealSeivBC4H3pmZRyPi4/zyUk4n0aEtZ2h/ZkPmXmAvwMjISI6OjjYYYmetVouF7L9ULJc6wFqWouVSB3S/lj179nTtWHM1Ojq6KPPS5Jr+SeBkZh4t64dovwh8t1y2ofz7xJT+66fsvw54fIZ2SVKPzBr6mfm/wGMRMfl7zhbgW8BhYPIOnJ3AHWX5MPCWchfPZuBMuQz0ReCKiLi4vIF7RWmTJPVIk8s7AO8EPhsR5wOPAG+j/YJxe0RcDzwKXFf63gVcDYwBPyl9ycxTEfFB4Cul3wcy81RXqpAkNdIo9DPzQWCkw6YtHfomcMM0x7kFuGUuA5QkdY+fyJWkihj6klQRQ1+SKmLoS1JFDH1JqoihL0kVMfQlqSKGviRVxNCXpIoY+pJUEUNfkipi6EtSRQx9SaqIoS9JFTH0Jakihr4kVcTQl6SKGPqSVBFDX5IqYuhLUkUMfUmqiKEvSRUx9CWpIoa+JFXE0Jekihj6klQRQ1+SKmLoS1JFDH1JqoihL0kVMfQlqSKGviRVxNCXpIoY+pJUEUNfkipi6EtSRVb2ewCLaXx8nD179vT8cfvxmJLUhGf6klSRxqEfESsi4qsRcWdZvzQijkbEwxFxW0ScX9pXlfWxsn3DlGO8r7SfiIgru12MJGlmcznTvxE4PmX9I8DNmbkROA1cX9qvB05n5suAm0s/IuIyYDvwcmAr8MmIWLGw4UuS5qJR6EfEOuD3gL8v6wG8AThUuuwHrinL28o6ZfuW0n8bcDAzn8rM7wBjwKZuFCFJaqbpG7l/BfwZcFFZfzHwZGaeLesngbVleS3wGEBmno2IM6X/WuC+Kcecus/TImIXsAtgaGiIVqvVtJZnWbVqFcPDw/Pef74WMuZOJiYmun7MfrGWpWe51AHdr6Uf+TFpseZl1tCPiN8HnsjMYxExOtncoWvOsm2mfX7ZkLkX2AswMjKSo6Oj53Zp7MCBA5w4cWLe+8/Xjh07unq8VqvFQp6HpcRalp7lUgd0v5Z+3ok3Ojq6KPPS5Ez/tcCbIuJq4ALgBbTP/FdHxMpytr8OeLz0PwmsB05GxErghcCpKe2Tpu4jSeqBWa/pZ+b7MnNdZm6g/UbsPZn5h8C9wLWl207gjrJ8uKxTtt+TmVnat5e7ey4FNgJf7lolkqRZLeTDWe8BDkbEh4CvAvtK+z7gMxExRvsMfztAZj4UEbcD3wLOAjdk5i8W8PiSpDmaU+hnZgtoleVH6HD3TWb+DLhumv1vAm6a6yAlSd3hJ3IlqSKGviRVxNCXpIoY+pJUEUNfkipi6EtSRQx9SaqIoS9JFTH0Jakihr4kVcTQl6SKGPqSVBFDX5IqYuhLUkUMfUmqiKEvSRUx9CWpIoa+JFXE0Jekihj6klQRQ1+SKmLoS1JFDH1JqoihL0kVMfQlqSKGviRVxNCXpIoY+pJUEUNfkipi6EtSRQx9SarIyn4PQMvDnj17GvUbHh5u3LebjyupzTN9SaqIoS9JFTH0Jakihr4kVcQ3cqUB0+03r+fy5rpvnA8+z/QlqSKzhn5ErI+IeyPieEQ8FBE3lvYXRcSRiHi4/HtxaY+I+EREjEXE1yPi8inH2ln6PxwROxevLElSJ03O9M8Cf5qZvwlsBm6IiMuA9wJ3Z+ZG4O6yDnAVsLH87AI+Be0XCWA38GpgE7B78oVCktQbs4Z+Zo5n5gNl+UfAcWAtsA3YX7rtB64py9uAW7PtPmB1RKwBrgSOZOapzDwNHAG2drUaSdKMIjObd47YAHwJeAXwaGaunrLtdGZeHBF3Ah/OzP8o7XcD7wFGgQsy80Ol/c+Bn2bmR895jF20f0NgaGjoVQcPHpx3cadOneKpp56a9/7ztWbNmq4eb2JiggsvvLCrx+y28fHxRv1WrVrV1Tnp9nM9F/2al6bPdVNzmZN+Pt9NdHtOuv1cz8VFF10071pe//rXH8vMkU7bGt+9ExEXAp8D3p2ZP4yIabt2aMsZ2p/ZkLkX2AswMjKSo6OjTYf4LAcOHODEiRPz3n++duzY0dXjtVotFvI89MJcvoahm3PS7ed6Lvo1L4tx907TOenn891Et+ekn3crjY6OLsp/X43u3omI82gH/mcz8/Ol+bvlsg3l3ydK+0lg/ZTd1wGPz9AuSeqRJnfvBLAPOJ6ZH5uy6TAweQfOTuCOKe1vKXfxbAbOZOY48EXgioi4uLyBe0VpkyT1SJPLO68F3gx8IyIeLG3vBz4M3B4R1wOPAteVbXcBVwNjwE+AtwFk5qmI+CDwldLvA5l5qitVSJIamTX0yxuy013A39KhfwI3THOsW4Bb5jJASVL3+IlcSaqIoS9JFTH0Jakihr4kVcTQl6SKGPqSVBFDX5IqYuhLUkUMfUmqiKEvSRUx9CWpIoa+JFXE0Jekihj6klQRQ1+SKmLoS1JFDH1JqoihL0kVMfQlqSKGviRVxNCXpIoY+pJUEUNfkipi6EtSRQx9SaqIoS9JFTH0Jakihr4kVcTQl6SKGPqSVBFDX5IqYuhLUkUMfUmqiKEvSRUx9CWpIoa+JFXE0Jekihj6klSRnod+RGyNiBMRMRYR7+3140tSzXoa+hGxAvhb4CrgMmBHRFzWyzFIUs16faa/CRjLzEcy8+fAQWBbj8cgSdWKzOzdg0VcC2zNzLeX9TcDr87Md0zpswvYVVaHgRMLeMhLgO8vYP+lYrnUAdayFC2XOsBaJv16Zr6k04aV8x/PvESHtme86mTmXmBvVx4s4v7MHOnGsfppudQB1rIULZc6wFqa6PXlnZPA+inr64DHezwGSapWr0P/K8DGiLg0Is4HtgOHezwGSapWTy/vZObZiHgH8EVgBXBLZj60iA/ZlctES8ByqQOsZSlaLnWAtcyqp2/kSpL6y0/kSlJFDH1JqsjAh/5sX+sQEasi4ray/WhEbOj9KJtpUMtbI+J7EfFg+Xl7P8Y5m4i4JSKeiIhvTrM9IuITpc6vR8TlvR5jUw1qGY2IM1Pm5C96PcYmImJ9RNwbEccj4qGIuLFDn4GYl4a1DMq8XBARX46Ir5Va/rJDn+5mWGYO7A/tN4P/C3gpcD7wNeCyc/r8MfDpsrwduK3f415ALW8F/qbfY21Qy+uAy4FvTrP9auALtD+3sRk42u8xL6CWUeDOfo+zQR1rgMvL8kXAf3b472sg5qVhLYMyLwFcWJbPA44Cm8/p09UMG/Qz/SZf67AN2F+WDwFbIqLTh8T6bdl8RUVmfgk4NUOXbcCt2XYfsDoi1vRmdHPToJaBkJnjmflAWf4RcBxYe063gZiXhrUMhPJcT5TV88rPuXfXdDXDBj301wKPTVk/ybMn/+k+mXkWOAO8uCejm5smtQD8QfnV+1BErO+wfRA0rXVQvKb8ev6FiHh5vwczm3J54Hdon1VONXDzMkMtMCDzEhErIuJB4AngSGZOOy/dyLBBD/1Zv9ahYZ+loMk4/xnYkJm/Bfw7v3z1HzSDMidNPED7e05eCfw18E99Hs+MIuJC4HPAuzPzh+du7rDLkp2XWWoZmHnJzF9k5m/T/oaCTRHxinO6dHVeBj30m3ytw9N9ImIl8EKW5q/rs9aSmT/IzKfK6t8Br+rR2Lpt2XwdR2b+cPLX88y8CzgvIi7p87A6iojzaIfkZzPz8x26DMy8zFbLIM3LpMx8EmgBW8/Z1NUMG/TQb/K1DoeBnWX5WuCeLO+ILDGz1nLO9dU30b6WOYgOA28pd4tsBs5k5ni/BzUfEfGrk9dXI2IT7f+nftDfUT1bGeM+4HhmfmyabgMxL01qGaB5eUlErC7LzwXeCHz7nG5dzbBef8tmV+U0X+sQER8A7s/Mw7T/4/hMRIzRfnXc3r8RT69hLe+KiDcBZ2nX8ta+DXgGEXGA9t0Tl0TESWA37TeoyMxPA3fRvlNkDPgJ8Lb+jHR2DWq5FvijiDgL/BTYvkRPKl4LvBn4Rrl+DPB+4Ndg4OalSS2DMi9rgP3R/gNTzwFuz8w7FzPD/BoGSarIoF/ekSTNgaEvSRUx9CWpIoa+JFXE0Jekihj6klQRQ1+SKvL/JGtN2MsF8TMAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "lgb_model = Lgb_Model(new_train, new_test, features, categoricals=cat_features)\n",
    "xgb_model = Xgb_Model(new_train, new_test, features, categoricals=cat_features)\n",
    "cat_model = Catb_Model(new_train, new_test, features, categoricals=cat_features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(17692, 398)\n",
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense (Dense)                (None, 200)               79800     \n",
      "_________________________________________________________________\n",
      "layer_normalization (LayerNo (None, 200)               400       \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 200)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 100)               20100     \n",
      "_________________________________________________________________\n",
      "layer_normalization_1 (Layer (None, 100)               200       \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 100)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 50)                5050      \n",
      "_________________________________________________________________\n",
      "layer_normalization_2 (Layer (None, 50)                100       \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 50)                0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 25)                1275      \n",
      "_________________________________________________________________\n",
      "layer_normalization_3 (Layer (None, 25)                50        \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 25)                0         \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 1)                 26        \n",
      "=================================================================\n",
      "Total params: 107,001\n",
      "Trainable params: 107,001\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Train on 14153 samples, validate on 3539 samples\n",
      "Epoch 1/100\n",
      "14048/14153 [============================>.] - ETA: 0s - loss: 1.7471\n",
      "Epoch 00001: val_loss improved from inf to 1.07020, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 2s 152us/sample - loss: 1.7449 - val_loss: 1.0702\n",
      "Epoch 2/100\n",
      "13472/14153 [===========================>..] - ETA: 0s - loss: 1.3998\n",
      "Epoch 00002: val_loss improved from 1.07020 to 1.02548, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 1s 98us/sample - loss: 1.3972 - val_loss: 1.0255\n",
      "Epoch 3/100\n",
      "13792/14153 [============================>.] - ETA: 0s - loss: 1.3097\n",
      "Epoch 00003: val_loss improved from 1.02548 to 1.01203, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 1s 102us/sample - loss: 1.3099 - val_loss: 1.0120\n",
      "Epoch 4/100\n",
      "13920/14153 [============================>.] - ETA: 0s - loss: 1.2340\n",
      "Epoch 00004: val_loss improved from 1.01203 to 0.99059, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 1s 99us/sample - loss: 1.2355 - val_loss: 0.9906\n",
      "Epoch 5/100\n",
      "14112/14153 [============================>.] - ETA: 0s - loss: 1.2053\n",
      "Epoch 00005: val_loss improved from 0.99059 to 0.98696, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 1s 98us/sample - loss: 1.2062 - val_loss: 0.9870\n",
      "Epoch 6/100\n",
      "13984/14153 [============================>.] - ETA: 0s - loss: 1.1748\n",
      "Epoch 00006: val_loss did not improve from 0.98696\n",
      "14153/14153 [==============================] - 1s 89us/sample - loss: 1.1750 - val_loss: 0.9971\n",
      "Epoch 7/100\n",
      "13536/14153 [===========================>..] - ETA: 0s - loss: 1.1704\n",
      "Epoch 00007: val_loss improved from 0.98696 to 0.98528, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 1s 99us/sample - loss: 1.1692 - val_loss: 0.9853\n",
      "Epoch 8/100\n",
      "13536/14153 [===========================>..] - ETA: 0s - loss: 1.1375\n",
      "Epoch 00008: val_loss improved from 0.98528 to 0.98122, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 2s 114us/sample - loss: 1.1372 - val_loss: 0.9812\n",
      "Epoch 9/100\n",
      "13856/14153 [============================>.] - ETA: 0s - loss: 1.1117\n",
      "Epoch 00009: val_loss improved from 0.98122 to 0.97691, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 2s 108us/sample - loss: 1.1122 - val_loss: 0.9769\n",
      "Epoch 10/100\n",
      "13984/14153 [============================>.] - ETA: 0s - loss: 1.1152\n",
      "Epoch 00010: val_loss improved from 0.97691 to 0.96635, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 2s 127us/sample - loss: 1.1152 - val_loss: 0.9663\n",
      "Epoch 11/100\n",
      "13600/14153 [===========================>..] - ETA: 0s - loss: 1.0902\n",
      "Epoch 00011: val_loss did not improve from 0.96635\n",
      "14153/14153 [==============================] - 1s 103us/sample - loss: 1.0910 - val_loss: 0.9796\n",
      "Epoch 12/100\n",
      "13984/14153 [============================>.] - ETA: 0s - loss: 1.0889\n",
      "Epoch 00012: val_loss improved from 0.96635 to 0.96421, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 1s 101us/sample - loss: 1.0890 - val_loss: 0.9642\n",
      "Epoch 13/100\n",
      "13920/14153 [============================>.] - ETA: 0s - loss: 1.0759\n",
      "Epoch 00013: val_loss improved from 0.96421 to 0.95815, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 1s 98us/sample - loss: 1.0746 - val_loss: 0.9582\n",
      "Epoch 14/100\n",
      "13920/14153 [============================>.] - ETA: 0s - loss: 1.0521\n",
      "Epoch 00014: val_loss improved from 0.95815 to 0.95569, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 1s 104us/sample - loss: 1.0516 - val_loss: 0.9557\n",
      "Epoch 15/100\n",
      "13920/14153 [============================>.] - ETA: 0s - loss: 1.0468\n",
      "Epoch 00015: val_loss did not improve from 0.95569\n",
      "14153/14153 [==============================] - 1s 101us/sample - loss: 1.0455 - val_loss: 0.9578\n",
      "Epoch 16/100\n",
      "13728/14153 [============================>.] - ETA: 0s - loss: 1.0404\n",
      "Epoch 00016: val_loss improved from 0.95569 to 0.94672, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 1s 101us/sample - loss: 1.0394 - val_loss: 0.9467\n",
      "Epoch 17/100\n",
      "13888/14153 [============================>.] - ETA: 0s - loss: 1.0188\n",
      "Epoch 00017: val_loss did not improve from 0.94672\n",
      "14153/14153 [==============================] - 1s 89us/sample - loss: 1.0166 - val_loss: 0.9577\n",
      "Epoch 18/100\n",
      "13984/14153 [============================>.] - ETA: 0s - loss: 1.0170\n",
      "Epoch 00018: val_loss improved from 0.94672 to 0.94324, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 1s 97us/sample - loss: 1.0170 - val_loss: 0.9432\n",
      "Epoch 19/100\n",
      "13824/14153 [============================>.] - ETA: 0s - loss: 1.0113\n",
      "Epoch 00019: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 1s 89us/sample - loss: 1.0118 - val_loss: 0.9452\n",
      "Epoch 20/100\n",
      "13952/14153 [============================>.] - ETA: 0s - loss: 0.9938\n",
      "Epoch 00020: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 1s 89us/sample - loss: 0.9956 - val_loss: 0.9555\n",
      "Epoch 21/100\n",
      "13696/14153 [============================>.] - ETA: 0s - loss: 0.9912\n",
      "Epoch 00021: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 1s 100us/sample - loss: 0.9926 - val_loss: 0.9800\n",
      "Epoch 22/100\n",
      "13952/14153 [============================>.] - ETA: 0s - loss: 0.9751\n",
      "Epoch 00022: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 2s 109us/sample - loss: 0.9769 - val_loss: 0.9634\n",
      "Epoch 23/100\n",
      "13824/14153 [============================>.] - ETA: 0s - loss: 0.9659\n",
      "Epoch 00023: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 1s 100us/sample - loss: 0.9654 - val_loss: 0.9566\n",
      "Epoch 24/100\n",
      "14048/14153 [============================>.] - ETA: 0s - loss: 0.9773\n",
      "Epoch 00024: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 1s 87us/sample - loss: 0.9770 - val_loss: 0.9578\n",
      "Epoch 25/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "13792/14153 [============================>.] - ETA: 0s - loss: 0.9627\n",
      "Epoch 00025: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 1s 85us/sample - loss: 0.9626 - val_loss: 0.9561\n",
      "Epoch 26/100\n",
      "13728/14153 [============================>.] - ETA: 0s - loss: 0.9389\n",
      "Epoch 00026: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 1s 85us/sample - loss: 0.9394 - val_loss: 0.9589\n",
      "Epoch 27/100\n",
      "13760/14153 [============================>.] - ETA: 0s - loss: 0.9403\n",
      "Epoch 00027: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 1s 85us/sample - loss: 0.9412 - val_loss: 0.9608\n",
      "Epoch 28/100\n",
      "13824/14153 [============================>.] - ETA: 0s - loss: 0.9442\n",
      "Epoch 00028: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 1s 93us/sample - loss: 0.9426 - val_loss: 0.9751\n",
      "Epoch 29/100\n",
      "13792/14153 [============================>.] - ETA: 0s - loss: 0.9354\n",
      "Epoch 00029: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 1s 89us/sample - loss: 0.9346 - val_loss: 0.9912\n",
      "Epoch 30/100\n",
      "13760/14153 [============================>.] - ETA: 0s - loss: 0.9201\n",
      "Epoch 00030: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 1s 85us/sample - loss: 0.9209 - val_loss: 0.9850\n",
      "Epoch 31/100\n",
      "13536/14153 [===========================>..] - ETA: 0s - loss: 0.9038\n",
      "Epoch 00031: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 1s 90us/sample - loss: 0.9115 - val_loss: 0.9837\n",
      "Epoch 32/100\n",
      "13888/14153 [============================>.] - ETA: 0s - loss: 0.9079\n",
      "Epoch 00032: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 1s 104us/sample - loss: 0.9061 - val_loss: 0.9794\n",
      "Epoch 33/100\n",
      "14112/14153 [============================>.] - ETA: 0s - loss: 0.8966\n",
      "Epoch 00033: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 2s 107us/sample - loss: 0.8970 - val_loss: 0.9913\n",
      "Epoch 34/100\n",
      "14080/14153 [============================>.] - ETA: 0s - loss: 0.8919\n",
      "Epoch 00034: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 1s 99us/sample - loss: 0.8933 - val_loss: 0.9721\n",
      "Epoch 35/100\n",
      "13888/14153 [============================>.] - ETA: 0s - loss: 0.8815\n",
      "Epoch 00035: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 2s 111us/sample - loss: 0.8827 - val_loss: 0.9949\n",
      "Epoch 36/100\n",
      "13696/14153 [============================>.] - ETA: 0s - loss: 0.8792\n",
      "Epoch 00036: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 1s 103us/sample - loss: 0.8763 - val_loss: 0.9954\n",
      "Epoch 37/100\n",
      "13920/14153 [============================>.] - ETA: 0s - loss: 0.8754\n",
      "Epoch 00037: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 1s 104us/sample - loss: 0.8763 - val_loss: 0.9881\n",
      "Epoch 38/100\n",
      "13984/14153 [============================>.] - ETA: 0s - loss: 0.8697\n",
      "Epoch 00038: val_loss did not improve from 0.94324\n",
      "14153/14153 [==============================] - 2s 113us/sample - loss: 0.8680 - val_loss: 1.0063\n",
      "------------------------------\n",
      "Partial score of fold 0 is: 0.6217678252863768\n",
      "------------------------------\n",
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_5 (Dense)              (None, 200)               79800     \n",
      "_________________________________________________________________\n",
      "layer_normalization_4 (Layer (None, 200)               400       \n",
      "_________________________________________________________________\n",
      "dropout_4 (Dropout)          (None, 200)               0         \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 100)               20100     \n",
      "_________________________________________________________________\n",
      "layer_normalization_5 (Layer (None, 100)               200       \n",
      "_________________________________________________________________\n",
      "dropout_5 (Dropout)          (None, 100)               0         \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 50)                5050      \n",
      "_________________________________________________________________\n",
      "layer_normalization_6 (Layer (None, 50)                100       \n",
      "_________________________________________________________________\n",
      "dropout_6 (Dropout)          (None, 50)                0         \n",
      "_________________________________________________________________\n",
      "dense_8 (Dense)              (None, 25)                1275      \n",
      "_________________________________________________________________\n",
      "layer_normalization_7 (Layer (None, 25)                50        \n",
      "_________________________________________________________________\n",
      "dropout_7 (Dropout)          (None, 25)                0         \n",
      "_________________________________________________________________\n",
      "dense_9 (Dense)              (None, 1)                 26        \n",
      "=================================================================\n",
      "Total params: 107,001\n",
      "Trainable params: 107,001\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Train on 14153 samples, validate on 3539 samples\n",
      "Epoch 1/100\n",
      "13824/14153 [============================>.] - ETA: 0s - loss: 1.9724\n",
      "Epoch 00001: val_loss improved from inf to 1.19552, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 3s 226us/sample - loss: 1.9601 - val_loss: 1.1955\n",
      "Epoch 2/100\n",
      "14112/14153 [============================>.] - ETA: 0s - loss: 1.4030\n",
      "Epoch 00002: val_loss improved from 1.19552 to 1.14775, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 2s 142us/sample - loss: 1.4026 - val_loss: 1.1477\n",
      "Epoch 3/100\n",
      "14112/14153 [============================>.] - ETA: 0s - loss: 1.2880\n",
      "Epoch 00003: val_loss improved from 1.14775 to 1.12608, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 2s 125us/sample - loss: 1.2877 - val_loss: 1.1261\n",
      "Epoch 4/100\n",
      "13696/14153 [============================>.] - ETA: 0s - loss: 1.2382\n",
      "Epoch 00004: val_loss improved from 1.12608 to 1.11432, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 2s 121us/sample - loss: 1.2390 - val_loss: 1.1143\n",
      "Epoch 5/100\n",
      "14016/14153 [============================>.] - ETA: 0s - loss: 1.1814\n",
      "Epoch 00005: val_loss improved from 1.11432 to 1.11425, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 2s 111us/sample - loss: 1.1816 - val_loss: 1.1143\n",
      "Epoch 6/100\n",
      "14144/14153 [============================>.] - ETA: 0s - loss: 1.1640\n",
      "Epoch 00006: val_loss improved from 1.11425 to 1.10229, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 2s 132us/sample - loss: 1.1638 - val_loss: 1.1023\n",
      "Epoch 7/100\n",
      "13600/14153 [===========================>..] - ETA: 0s - loss: 1.1462\n",
      "Epoch 00007: val_loss improved from 1.10229 to 1.09663, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 1s 101us/sample - loss: 1.1478 - val_loss: 1.0966\n",
      "Epoch 8/100\n",
      "14080/14153 [============================>.] - ETA: 0s - loss: 1.1219\n",
      "Epoch 00008: val_loss did not improve from 1.09663\n",
      "14153/14153 [==============================] - 1s 91us/sample - loss: 1.1233 - val_loss: 1.1093\n",
      "Epoch 9/100\n",
      "13920/14153 [============================>.] - ETA: 0s - loss: 1.1106\n",
      "Epoch 00009: val_loss improved from 1.09663 to 1.09344, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 1s 101us/sample - loss: 1.1097 - val_loss: 1.0934\n",
      "Epoch 10/100\n",
      "14112/14153 [============================>.] - ETA: 0s - loss: 1.0939\n",
      "Epoch 00010: val_loss did not improve from 1.09344\n",
      "14153/14153 [==============================] - 1s 104us/sample - loss: 1.0929 - val_loss: 1.1144\n",
      "Epoch 11/100\n",
      "13696/14153 [============================>.] - ETA: 0s - loss: 1.0836\n",
      "Epoch 00011: val_loss improved from 1.09344 to 1.09155, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 1s 101us/sample - loss: 1.0817 - val_loss: 1.0915\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 12/100\n",
      "13632/14153 [===========================>..] - ETA: 0s - loss: 1.0663\n",
      "Epoch 00012: val_loss did not improve from 1.09155\n",
      "14153/14153 [==============================] - 1s 90us/sample - loss: 1.0649 - val_loss: 1.1176\n",
      "Epoch 13/100\n",
      "13760/14153 [============================>.] - ETA: 0s - loss: 1.0472\n",
      "Epoch 00013: val_loss improved from 1.09155 to 1.06948, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 1s 106us/sample - loss: 1.0432 - val_loss: 1.0695\n",
      "Epoch 14/100\n",
      "14112/14153 [============================>.] - ETA: 0s - loss: 1.0449\n",
      "Epoch 00014: val_loss did not improve from 1.06948\n",
      "14153/14153 [==============================] - 1s 100us/sample - loss: 1.0443 - val_loss: 1.0734\n",
      "Epoch 15/100\n",
      "14048/14153 [============================>.] - ETA: 0s - loss: 1.0244\n",
      "Epoch 00015: val_loss did not improve from 1.06948\n",
      "14153/14153 [==============================] - 2s 107us/sample - loss: 1.0250 - val_loss: 1.0834\n",
      "Epoch 16/100\n",
      "13824/14153 [============================>.] - ETA: 0s - loss: 1.0222\n",
      "Epoch 00016: val_loss did not improve from 1.06948\n",
      "14153/14153 [==============================] - 1s 92us/sample - loss: 1.0245 - val_loss: 1.0732\n",
      "Epoch 17/100\n",
      "13664/14153 [===========================>..] - ETA: 0s - loss: 1.0000\n",
      "Epoch 00017: val_loss improved from 1.06948 to 1.05960, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 1s 100us/sample - loss: 1.0033 - val_loss: 1.0596\n",
      "Epoch 18/100\n",
      "13792/14153 [============================>.] - ETA: 0s - loss: 0.9952\n",
      "Epoch 00018: val_loss did not improve from 1.05960\n",
      "14153/14153 [==============================] - 1s 93us/sample - loss: 0.9922 - val_loss: 1.0786\n",
      "Epoch 19/100\n",
      "14016/14153 [============================>.] - ETA: 0s - loss: 0.9843\n",
      "Epoch 00019: val_loss did not improve from 1.05960\n",
      "14153/14153 [==============================] - 1s 93us/sample - loss: 0.9858 - val_loss: 1.0826\n",
      "Epoch 20/100\n",
      "13600/14153 [===========================>..] - ETA: 0s - loss: 0.9826\n",
      "Epoch 00020: val_loss improved from 1.05960 to 1.05358, saving model to nn_model.w8\n",
      "14153/14153 [==============================] - 2s 113us/sample - loss: 0.9831 - val_loss: 1.0536\n",
      "Epoch 21/100\n",
      "14048/14153 [============================>.] - ETA: 0s - loss: 0.9628\n",
      "Epoch 00021: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 104us/sample - loss: 0.9631 - val_loss: 1.0644\n",
      "Epoch 22/100\n",
      "13856/14153 [============================>.] - ETA: 0s - loss: 0.9669\n",
      "Epoch 00022: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 97us/sample - loss: 0.9669 - val_loss: 1.0763\n",
      "Epoch 23/100\n",
      "14048/14153 [============================>.] - ETA: 0s - loss: 0.9555\n",
      "Epoch 00023: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 2s 107us/sample - loss: 0.9559 - val_loss: 1.0544\n",
      "Epoch 24/100\n",
      "13536/14153 [===========================>..] - ETA: 0s - loss: 0.9347\n",
      "Epoch 00024: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 91us/sample - loss: 0.9355 - val_loss: 1.0749\n",
      "Epoch 25/100\n",
      "13952/14153 [============================>.] - ETA: 0s - loss: 0.9268\n",
      "Epoch 00025: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 89us/sample - loss: 0.9279 - val_loss: 1.0647\n",
      "Epoch 26/100\n",
      "13664/14153 [===========================>..] - ETA: 0s - loss: 0.9326\n",
      "Epoch 00026: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 89us/sample - loss: 0.9312 - val_loss: 1.0759\n",
      "Epoch 27/100\n",
      "13536/14153 [===========================>..] - ETA: 0s - loss: 0.9229\n",
      "Epoch 00027: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 90us/sample - loss: 0.9259 - val_loss: 1.0643\n",
      "Epoch 28/100\n",
      "14016/14153 [============================>.] - ETA: 0s - loss: 0.9146\n",
      "Epoch 00028: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 99us/sample - loss: 0.9137 - val_loss: 1.0983\n",
      "Epoch 29/100\n",
      "13728/14153 [============================>.] - ETA: 0s - loss: 0.8989\n",
      "Epoch 00029: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 94us/sample - loss: 0.9025 - val_loss: 1.0654\n",
      "Epoch 30/100\n",
      "13824/14153 [============================>.] - ETA: 0s - loss: 0.9058\n",
      "Epoch 00030: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 90us/sample - loss: 0.9028 - val_loss: 1.0751\n",
      "Epoch 31/100\n",
      "13760/14153 [============================>.] - ETA: 0s - loss: 0.8937\n",
      "Epoch 00031: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 89us/sample - loss: 0.8936 - val_loss: 1.0720\n",
      "Epoch 32/100\n",
      "13952/14153 [============================>.] - ETA: 0s - loss: 0.8973\n",
      "Epoch 00032: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 93us/sample - loss: 0.8959 - val_loss: 1.0836\n",
      "Epoch 33/100\n",
      "13856/14153 [============================>.] - ETA: 0s - loss: 0.8783\n",
      "Epoch 00033: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 97us/sample - loss: 0.8777 - val_loss: 1.1010\n",
      "Epoch 34/100\n",
      "13632/14153 [===========================>..] - ETA: 0s - loss: 0.8710\n",
      "Epoch 00034: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 106us/sample - loss: 0.8719 - val_loss: 1.0924\n",
      "Epoch 35/100\n",
      "13792/14153 [============================>.] - ETA: 0s - loss: 0.8701\n",
      "Epoch 00035: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 97us/sample - loss: 0.8694 - val_loss: 1.1043\n",
      "Epoch 36/100\n",
      "13760/14153 [============================>.] - ETA: 0s - loss: 0.8589\n",
      "Epoch 00036: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 97us/sample - loss: 0.8572 - val_loss: 1.0812\n",
      "Epoch 37/100\n",
      "13504/14153 [===========================>..] - ETA: 0s - loss: 0.8600\n",
      "Epoch 00037: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 91us/sample - loss: 0.8619 - val_loss: 1.0725\n",
      "Epoch 38/100\n",
      "13600/14153 [===========================>..] - ETA: 0s - loss: 0.8516\n",
      "Epoch 00038: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 90us/sample - loss: 0.8484 - val_loss: 1.1101\n",
      "Epoch 39/100\n",
      "13792/14153 [============================>.] - ETA: 0s - loss: 0.8476\n",
      "Epoch 00039: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 88us/sample - loss: 0.8487 - val_loss: 1.0984\n",
      "Epoch 40/100\n",
      "13824/14153 [============================>.] - ETA: 0s - loss: 0.8490\n",
      "Epoch 00040: val_loss did not improve from 1.05358\n",
      "14153/14153 [==============================] - 1s 97us/sample - loss: 0.8469 - val_loss: 1.0974\n",
      "------------------------------\n",
      "Partial score of fold 1 is: 0.563515882904096\n",
      "------------------------------\n",
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_10 (Dense)             (None, 200)               79800     \n",
      "_________________________________________________________________\n",
      "layer_normalization_8 (Layer (None, 200)               400       \n",
      "_________________________________________________________________\n",
      "dropout_8 (Dropout)          (None, 200)               0         \n",
      "_________________________________________________________________\n",
      "dense_11 (Dense)             (None, 100)               20100     \n",
      "_________________________________________________________________\n",
      "layer_normalization_9 (Layer (None, 100)               200       \n",
      "_________________________________________________________________\n",
      "dropout_9 (Dropout)          (None, 100)               0         \n",
      "_________________________________________________________________\n",
      "dense_12 (Dense)             (None, 50)                5050      \n",
      "_________________________________________________________________\n",
      "layer_normalization_10 (Laye (None, 50)                100       \n",
      "_________________________________________________________________\n",
      "dropout_10 (Dropout)         (None, 50)                0         \n",
      "_________________________________________________________________\n",
      "dense_13 (Dense)             (None, 25)                1275      \n",
      "_________________________________________________________________\n",
      "layer_normalization_11 (Laye (None, 25)                50        \n",
      "_________________________________________________________________\n",
      "dropout_11 (Dropout)         (None, 25)                0         \n",
      "_________________________________________________________________\n",
      "dense_14 (Dense)             (None, 1)                 26        \n",
      "=================================================================\n",
      "Total params: 107,001\n",
      "Trainable params: 107,001\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Train on 14154 samples, validate on 3538 samples\n",
      "Epoch 1/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "14144/14154 [============================>.] - ETA: 0s - loss: 2.0089\n",
      "Epoch 00001: val_loss improved from inf to 1.10416, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 154us/sample - loss: 2.0090 - val_loss: 1.1042\n",
      "Epoch 2/100\n",
      "13632/14154 [===========================>..] - ETA: 0s - loss: 1.4488\n",
      "Epoch 00002: val_loss improved from 1.10416 to 1.08458, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 119us/sample - loss: 1.4471 - val_loss: 1.0846\n",
      "Epoch 3/100\n",
      "14144/14154 [============================>.] - ETA: 0s - loss: 1.3257\n",
      "Epoch 00003: val_loss improved from 1.08458 to 1.06838, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 112us/sample - loss: 1.3257 - val_loss: 1.0684\n",
      "Epoch 4/100\n",
      "13792/14154 [============================>.] - ETA: 0s - loss: 1.2603\n",
      "Epoch 00004: val_loss improved from 1.06838 to 1.05207, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 112us/sample - loss: 1.2588 - val_loss: 1.0521\n",
      "Epoch 5/100\n",
      "13856/14154 [============================>.] - ETA: 0s - loss: 1.2180\n",
      "Epoch 00005: val_loss improved from 1.05207 to 1.03797, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 112us/sample - loss: 1.2152 - val_loss: 1.0380\n",
      "Epoch 6/100\n",
      "13760/14154 [============================>.] - ETA: 0s - loss: 1.1775\n",
      "Epoch 00006: val_loss did not improve from 1.03797\n",
      "14154/14154 [==============================] - 2s 107us/sample - loss: 1.1782 - val_loss: 1.0757\n",
      "Epoch 7/100\n",
      "13920/14154 [============================>.] - ETA: 0s - loss: 1.1612\n",
      "Epoch 00007: val_loss did not improve from 1.03797\n",
      "14154/14154 [==============================] - 1s 92us/sample - loss: 1.1622 - val_loss: 1.0643\n",
      "Epoch 8/100\n",
      "13632/14154 [===========================>..] - ETA: 0s - loss: 1.1253\n",
      "Epoch 00008: val_loss did not improve from 1.03797\n",
      "14154/14154 [==============================] - 1s 90us/sample - loss: 1.1219 - val_loss: 1.0444\n",
      "Epoch 9/100\n",
      "13856/14154 [============================>.] - ETA: 0s - loss: 1.1238\n",
      "Epoch 00009: val_loss improved from 1.03797 to 1.01856, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 1s 100us/sample - loss: 1.1236 - val_loss: 1.0186\n",
      "Epoch 10/100\n",
      "13920/14154 [============================>.] - ETA: 0s - loss: 1.0923\n",
      "Epoch 00010: val_loss did not improve from 1.01856\n",
      "14154/14154 [==============================] - 1s 92us/sample - loss: 1.0955 - val_loss: 1.0193\n",
      "Epoch 11/100\n",
      "13824/14154 [============================>.] - ETA: 0s - loss: 1.0823\n",
      "Epoch 00011: val_loss improved from 1.01856 to 1.01073, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 1s 101us/sample - loss: 1.0818 - val_loss: 1.0107\n",
      "Epoch 12/100\n",
      "13760/14154 [============================>.] - ETA: 0s - loss: 1.0734\n",
      "Epoch 00012: val_loss did not improve from 1.01073\n",
      "14154/14154 [==============================] - 1s 105us/sample - loss: 1.0698 - val_loss: 1.0194\n",
      "Epoch 13/100\n",
      "13856/14154 [============================>.] - ETA: 0s - loss: 1.0620\n",
      "Epoch 00013: val_loss did not improve from 1.01073\n",
      "14154/14154 [==============================] - 1s 97us/sample - loss: 1.0586 - val_loss: 1.0118\n",
      "Epoch 14/100\n",
      "13632/14154 [===========================>..] - ETA: 0s - loss: 1.0501\n",
      "Epoch 00014: val_loss did not improve from 1.01073\n",
      "14154/14154 [==============================] - 1s 90us/sample - loss: 1.0537 - val_loss: 1.0352\n",
      "Epoch 15/100\n",
      "13888/14154 [============================>.] - ETA: 0s - loss: 1.0399\n",
      "Epoch 00015: val_loss did not improve from 1.01073\n",
      "14154/14154 [==============================] - 1s 89us/sample - loss: 1.0374 - val_loss: 1.0307\n",
      "Epoch 16/100\n",
      "14080/14154 [============================>.] - ETA: 0s - loss: 1.0269\n",
      "Epoch 00016: val_loss improved from 1.01073 to 0.99996, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 1s 102us/sample - loss: 1.0267 - val_loss: 1.0000\n",
      "Epoch 17/100\n",
      "13856/14154 [============================>.] - ETA: 0s - loss: 1.0127\n",
      "Epoch 00017: val_loss did not improve from 0.99996\n",
      "14154/14154 [==============================] - 1s 93us/sample - loss: 1.0132 - val_loss: 1.0147\n",
      "Epoch 18/100\n",
      "13760/14154 [============================>.] - ETA: 0s - loss: 1.0037\n",
      "Epoch 00018: val_loss improved from 0.99996 to 0.99756, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 1s 96us/sample - loss: 1.0050 - val_loss: 0.9976\n",
      "Epoch 19/100\n",
      "13664/14154 [===========================>..] - ETA: 0s - loss: 0.9923\n",
      "Epoch 00019: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 90us/sample - loss: 0.9891 - val_loss: 1.0087\n",
      "Epoch 20/100\n",
      "13824/14154 [============================>.] - ETA: 0s - loss: 0.9809\n",
      "Epoch 00020: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 98us/sample - loss: 0.9811 - val_loss: 1.0245\n",
      "Epoch 21/100\n",
      "13920/14154 [============================>.] - ETA: 0s - loss: 0.9753\n",
      "Epoch 00021: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 92us/sample - loss: 0.9750 - val_loss: 1.0140\n",
      "Epoch 22/100\n",
      "14048/14154 [============================>.] - ETA: 0s - loss: 0.9713\n",
      "Epoch 00022: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 92us/sample - loss: 0.9713 - val_loss: 1.0090\n",
      "Epoch 23/100\n",
      "13664/14154 [===========================>..] - ETA: 0s - loss: 0.9575\n",
      "Epoch 00023: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 99us/sample - loss: 0.9580 - val_loss: 1.0224\n",
      "Epoch 24/100\n",
      "14144/14154 [============================>.] - ETA: 0s - loss: 0.9467\n",
      "Epoch 00024: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 93us/sample - loss: 0.9466 - val_loss: 1.0140\n",
      "Epoch 25/100\n",
      "14080/14154 [============================>.] - ETA: 0s - loss: 0.9353\n",
      "Epoch 00025: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 91us/sample - loss: 0.9356 - val_loss: 1.0220\n",
      "Epoch 26/100\n",
      "13984/14154 [============================>.] - ETA: 0s - loss: 0.9226\n",
      "Epoch 00026: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 92us/sample - loss: 0.9219 - val_loss: 1.0417\n",
      "Epoch 27/100\n",
      "13824/14154 [============================>.] - ETA: 0s - loss: 0.9320\n",
      "Epoch 00027: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 89us/sample - loss: 0.9329 - val_loss: 1.0066\n",
      "Epoch 28/100\n",
      "14016/14154 [============================>.] - ETA: 0s - loss: 0.9280\n",
      "Epoch 00028: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 93us/sample - loss: 0.9276 - val_loss: 1.0271\n",
      "Epoch 29/100\n",
      "14080/14154 [============================>.] - ETA: 0s - loss: 0.9159\n",
      "Epoch 00029: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 96us/sample - loss: 0.9159 - val_loss: 1.0365\n",
      "Epoch 30/100\n",
      "13728/14154 [============================>.] - ETA: 0s - loss: 0.9133\n",
      "Epoch 00030: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 98us/sample - loss: 0.9147 - val_loss: 1.0112\n",
      "Epoch 31/100\n",
      "14112/14154 [============================>.] - ETA: 0s - loss: 0.8984\n",
      "Epoch 00031: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 91us/sample - loss: 0.8976 - val_loss: 1.0177\n",
      "Epoch 32/100\n",
      "13600/14154 [===========================>..] - ETA: 0s - loss: 0.8964\n",
      "Epoch 00032: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 95us/sample - loss: 0.8942 - val_loss: 1.0231\n",
      "Epoch 33/100\n",
      "13920/14154 [============================>.] - ETA: 0s - loss: 0.8859\n",
      "Epoch 00033: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 2s 106us/sample - loss: 0.8858 - val_loss: 1.0310\n",
      "Epoch 34/100\n",
      "13568/14154 [===========================>..] - ETA: 0s - loss: 0.8818\n",
      "Epoch 00034: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 98us/sample - loss: 0.8806 - val_loss: 1.0272\n",
      "Epoch 35/100\n",
      "14048/14154 [============================>.] - ETA: 0s - loss: 0.8742\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 00035: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 104us/sample - loss: 0.8745 - val_loss: 1.0211\n",
      "Epoch 36/100\n",
      "14112/14154 [============================>.] - ETA: 0s - loss: 0.8698\n",
      "Epoch 00036: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 95us/sample - loss: 0.8699 - val_loss: 1.0373\n",
      "Epoch 37/100\n",
      "13664/14154 [===========================>..] - ETA: 0s - loss: 0.8543\n",
      "Epoch 00037: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 90us/sample - loss: 0.8588 - val_loss: 1.0555\n",
      "Epoch 38/100\n",
      "13920/14154 [============================>.] - ETA: 0s - loss: 0.8599\n",
      "Epoch 00038: val_loss did not improve from 0.99756\n",
      "14154/14154 [==============================] - 1s 89us/sample - loss: 0.8580 - val_loss: 1.0178\n",
      "------------------------------\n",
      "Partial score of fold 2 is: 0.5527325763915925\n",
      "------------------------------\n",
      "Model: \"sequential_3\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_15 (Dense)             (None, 200)               79800     \n",
      "_________________________________________________________________\n",
      "layer_normalization_12 (Laye (None, 200)               400       \n",
      "_________________________________________________________________\n",
      "dropout_12 (Dropout)         (None, 200)               0         \n",
      "_________________________________________________________________\n",
      "dense_16 (Dense)             (None, 100)               20100     \n",
      "_________________________________________________________________\n",
      "layer_normalization_13 (Laye (None, 100)               200       \n",
      "_________________________________________________________________\n",
      "dropout_13 (Dropout)         (None, 100)               0         \n",
      "_________________________________________________________________\n",
      "dense_17 (Dense)             (None, 50)                5050      \n",
      "_________________________________________________________________\n",
      "layer_normalization_14 (Laye (None, 50)                100       \n",
      "_________________________________________________________________\n",
      "dropout_14 (Dropout)         (None, 50)                0         \n",
      "_________________________________________________________________\n",
      "dense_18 (Dense)             (None, 25)                1275      \n",
      "_________________________________________________________________\n",
      "layer_normalization_15 (Laye (None, 25)                50        \n",
      "_________________________________________________________________\n",
      "dropout_15 (Dropout)         (None, 25)                0         \n",
      "_________________________________________________________________\n",
      "dense_19 (Dense)             (None, 1)                 26        \n",
      "=================================================================\n",
      "Total params: 107,001\n",
      "Trainable params: 107,001\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Train on 14154 samples, validate on 3538 samples\n",
      "Epoch 1/100\n",
      "13920/14154 [============================>.] - ETA: 0s - loss: 1.8301\n",
      "Epoch 00001: val_loss improved from inf to 1.15847, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 148us/sample - loss: 1.8211 - val_loss: 1.1585\n",
      "Epoch 2/100\n",
      "14016/14154 [============================>.] - ETA: 0s - loss: 1.3884\n",
      "Epoch 00002: val_loss improved from 1.15847 to 1.13715, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 111us/sample - loss: 1.3881 - val_loss: 1.1371\n",
      "Epoch 3/100\n",
      "13952/14154 [============================>.] - ETA: 0s - loss: 1.2694\n",
      "Epoch 00003: val_loss improved from 1.13715 to 1.11900, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 116us/sample - loss: 1.2724 - val_loss: 1.1190\n",
      "Epoch 4/100\n",
      "13856/14154 [============================>.] - ETA: 0s - loss: 1.2121\n",
      "Epoch 00004: val_loss improved from 1.11900 to 1.11692, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 109us/sample - loss: 1.2083 - val_loss: 1.1169\n",
      "Epoch 5/100\n",
      "14144/14154 [============================>.] - ETA: 0s - loss: 1.1882\n",
      "Epoch 00005: val_loss improved from 1.11692 to 1.09232, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 116us/sample - loss: 1.1886 - val_loss: 1.0923\n",
      "Epoch 6/100\n",
      "13664/14154 [===========================>..] - ETA: 0s - loss: 1.1458\n",
      "Epoch 00006: val_loss improved from 1.09232 to 1.07027, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 111us/sample - loss: 1.1424 - val_loss: 1.0703\n",
      "Epoch 7/100\n",
      "13920/14154 [============================>.] - ETA: 0s - loss: 1.1260\n",
      "Epoch 00007: val_loss did not improve from 1.07027\n",
      "14154/14154 [==============================] - 1s 96us/sample - loss: 1.1245 - val_loss: 1.1164\n",
      "Epoch 8/100\n",
      "13792/14154 [============================>.] - ETA: 0s - loss: 1.1084\n",
      "Epoch 00008: val_loss improved from 1.07027 to 1.07010, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 107us/sample - loss: 1.1068 - val_loss: 1.0701\n",
      "Epoch 9/100\n",
      "13824/14154 [============================>.] - ETA: 0s - loss: 1.1068\n",
      "Epoch 00009: val_loss improved from 1.07010 to 1.06861, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 117us/sample - loss: 1.1054 - val_loss: 1.0686\n",
      "Epoch 10/100\n",
      "13568/14154 [===========================>..] - ETA: 0s - loss: 1.0826\n",
      "Epoch 00010: val_loss improved from 1.06861 to 1.06563, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 1s 105us/sample - loss: 1.0847 - val_loss: 1.0656\n",
      "Epoch 11/100\n",
      "13792/14154 [============================>.] - ETA: 0s - loss: 1.0657\n",
      "Epoch 00011: val_loss improved from 1.06563 to 1.06271, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 108us/sample - loss: 1.0655 - val_loss: 1.0627\n",
      "Epoch 12/100\n",
      "13504/14154 [===========================>..] - ETA: 0s - loss: 1.0524\n",
      "Epoch 00012: val_loss did not improve from 1.06271\n",
      "14154/14154 [==============================] - 1s 86us/sample - loss: 1.0527 - val_loss: 1.0645\n",
      "Epoch 13/100\n",
      "13664/14154 [===========================>..] - ETA: 0s - loss: 1.0386\n",
      "Epoch 00013: val_loss improved from 1.06271 to 1.04409, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 1s 102us/sample - loss: 1.0366 - val_loss: 1.0441\n",
      "Epoch 14/100\n",
      "13984/14154 [============================>.] - ETA: 0s - loss: 1.0317\n",
      "Epoch 00014: val_loss improved from 1.04409 to 1.03402, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 116us/sample - loss: 1.0300 - val_loss: 1.0340\n",
      "Epoch 15/100\n",
      "14080/14154 [============================>.] - ETA: 0s - loss: 1.0251\n",
      "Epoch 00015: val_loss did not improve from 1.03402\n",
      "14154/14154 [==============================] - 1s 95us/sample - loss: 1.0252 - val_loss: 1.0375\n",
      "Epoch 16/100\n",
      "13952/14154 [============================>.] - ETA: 0s - loss: 1.0171\n",
      "Epoch 00016: val_loss did not improve from 1.03402\n",
      "14154/14154 [==============================] - 1s 99us/sample - loss: 1.0160 - val_loss: 1.0374\n",
      "Epoch 17/100\n",
      "13728/14154 [============================>.] - ETA: 0s - loss: 1.0103\n",
      "Epoch 00017: val_loss did not improve from 1.03402\n",
      "14154/14154 [==============================] - 1s 85us/sample - loss: 1.0079 - val_loss: 1.0626\n",
      "Epoch 18/100\n",
      "14080/14154 [============================>.] - ETA: 0s - loss: 0.9971\n",
      "Epoch 00018: val_loss did not improve from 1.03402\n",
      "14154/14154 [==============================] - 2s 108us/sample - loss: 0.9960 - val_loss: 1.0504\n",
      "Epoch 19/100\n",
      "13824/14154 [============================>.] - ETA: 0s - loss: 0.9890\n",
      "Epoch 00019: val_loss did not improve from 1.03402\n",
      "14154/14154 [==============================] - 1s 93us/sample - loss: 0.9879 - val_loss: 1.0441\n",
      "Epoch 20/100\n",
      "14016/14154 [============================>.] - ETA: 0s - loss: 0.9758\n",
      "Epoch 00020: val_loss improved from 1.03402 to 1.02576, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 1s 103us/sample - loss: 0.9743 - val_loss: 1.0258\n",
      "Epoch 21/100\n",
      "14112/14154 [============================>.] - ETA: 0s - loss: 0.9641\n",
      "Epoch 00021: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 95us/sample - loss: 0.9645 - val_loss: 1.0596\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 22/100\n",
      "13792/14154 [============================>.] - ETA: 0s - loss: 0.9590\n",
      "Epoch 00022: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 85us/sample - loss: 0.9582 - val_loss: 1.0522\n",
      "Epoch 23/100\n",
      "13600/14154 [===========================>..] - ETA: 0s - loss: 0.9423\n",
      "Epoch 00023: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 90us/sample - loss: 0.9464 - val_loss: 1.0438\n",
      "Epoch 24/100\n",
      "14112/14154 [============================>.] - ETA: 0s - loss: 0.9465\n",
      "Epoch 00024: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 94us/sample - loss: 0.9459 - val_loss: 1.0558\n",
      "Epoch 25/100\n",
      "14144/14154 [============================>.] - ETA: 0s - loss: 0.9339\n",
      "Epoch 00025: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 100us/sample - loss: 0.9342 - val_loss: 1.0531\n",
      "Epoch 26/100\n",
      "13920/14154 [============================>.] - ETA: 0s - loss: 0.9354\n",
      "Epoch 00026: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 96us/sample - loss: 0.9353 - val_loss: 1.0428\n",
      "Epoch 27/100\n",
      "13536/14154 [===========================>..] - ETA: 0s - loss: 0.9214\n",
      "Epoch 00027: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 87us/sample - loss: 0.9264 - val_loss: 1.0711\n",
      "Epoch 28/100\n",
      "13728/14154 [============================>.] - ETA: 0s - loss: 0.9143\n",
      "Epoch 00028: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 85us/sample - loss: 0.9131 - val_loss: 1.0693\n",
      "Epoch 29/100\n",
      "13632/14154 [===========================>..] - ETA: 0s - loss: 0.9094\n",
      "Epoch 00029: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 85us/sample - loss: 0.9089 - val_loss: 1.0766\n",
      "Epoch 30/100\n",
      "13760/14154 [============================>.] - ETA: 0s - loss: 0.8955\n",
      "Epoch 00030: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 85us/sample - loss: 0.8960 - val_loss: 1.0865\n",
      "Epoch 31/100\n",
      "13632/14154 [===========================>..] - ETA: 0s - loss: 0.8984\n",
      "Epoch 00031: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 86us/sample - loss: 0.8968 - val_loss: 1.1062\n",
      "Epoch 32/100\n",
      "13728/14154 [============================>.] - ETA: 0s - loss: 0.8832\n",
      "Epoch 00032: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 85us/sample - loss: 0.8824 - val_loss: 1.0782\n",
      "Epoch 33/100\n",
      "13728/14154 [============================>.] - ETA: 0s - loss: 0.8729\n",
      "Epoch 00033: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 87us/sample - loss: 0.8754 - val_loss: 1.0890\n",
      "Epoch 34/100\n",
      "14016/14154 [============================>.] - ETA: 0s - loss: 0.8711\n",
      "Epoch 00034: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 88us/sample - loss: 0.8723 - val_loss: 1.0646\n",
      "Epoch 35/100\n",
      "13856/14154 [============================>.] - ETA: 0s - loss: 0.8549\n",
      "Epoch 00035: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 92us/sample - loss: 0.8575 - val_loss: 1.0982\n",
      "Epoch 36/100\n",
      "13696/14154 [============================>.] - ETA: 0s - loss: 0.8562\n",
      "Epoch 00036: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 90us/sample - loss: 0.8553 - val_loss: 1.0887\n",
      "Epoch 37/100\n",
      "13888/14154 [============================>.] - ETA: 0s - loss: 0.8598\n",
      "Epoch 00037: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 90us/sample - loss: 0.8579 - val_loss: 1.1148\n",
      "Epoch 38/100\n",
      "13824/14154 [============================>.] - ETA: 0s - loss: 0.8558\n",
      "Epoch 00038: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 2s 106us/sample - loss: 0.8524 - val_loss: 1.1165\n",
      "Epoch 39/100\n",
      "13856/14154 [============================>.] - ETA: 0s - loss: 0.8411\n",
      "Epoch 00039: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 89us/sample - loss: 0.8415 - val_loss: 1.1073\n",
      "Epoch 40/100\n",
      "13920/14154 [============================>.] - ETA: 0s - loss: 0.8353\n",
      "Epoch 00040: val_loss did not improve from 1.02576\n",
      "14154/14154 [==============================] - 1s 92us/sample - loss: 0.8345 - val_loss: 1.1021\n",
      "------------------------------\n",
      "Partial score of fold 3 is: 0.5768284288543732\n",
      "------------------------------\n",
      "Model: \"sequential_4\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_20 (Dense)             (None, 200)               79800     \n",
      "_________________________________________________________________\n",
      "layer_normalization_16 (Laye (None, 200)               400       \n",
      "_________________________________________________________________\n",
      "dropout_16 (Dropout)         (None, 200)               0         \n",
      "_________________________________________________________________\n",
      "dense_21 (Dense)             (None, 100)               20100     \n",
      "_________________________________________________________________\n",
      "layer_normalization_17 (Laye (None, 100)               200       \n",
      "_________________________________________________________________\n",
      "dropout_17 (Dropout)         (None, 100)               0         \n",
      "_________________________________________________________________\n",
      "dense_22 (Dense)             (None, 50)                5050      \n",
      "_________________________________________________________________\n",
      "layer_normalization_18 (Laye (None, 50)                100       \n",
      "_________________________________________________________________\n",
      "dropout_18 (Dropout)         (None, 50)                0         \n",
      "_________________________________________________________________\n",
      "dense_23 (Dense)             (None, 25)                1275      \n",
      "_________________________________________________________________\n",
      "layer_normalization_19 (Laye (None, 25)                50        \n",
      "_________________________________________________________________\n",
      "dropout_19 (Dropout)         (None, 25)                0         \n",
      "_________________________________________________________________\n",
      "dense_24 (Dense)             (None, 1)                 26        \n",
      "=================================================================\n",
      "Total params: 107,001\n",
      "Trainable params: 107,001\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Train on 14154 samples, validate on 3538 samples\n",
      "Epoch 1/100\n",
      "13728/14154 [============================>.] - ETA: 0s - loss: 2.0288\n",
      "Epoch 00001: val_loss improved from inf to 1.18730, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 136us/sample - loss: 2.0167 - val_loss: 1.1873\n",
      "Epoch 2/100\n",
      "13632/14154 [===========================>..] - ETA: 0s - loss: 1.4467\n",
      "Epoch 00002: val_loss improved from 1.18730 to 1.17157, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 2s 107us/sample - loss: 1.4431 - val_loss: 1.1716\n",
      "Epoch 3/100\n",
      "13504/14154 [===========================>..] - ETA: 0s - loss: 1.3201\n",
      "Epoch 00003: val_loss improved from 1.17157 to 1.16167, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 1s 101us/sample - loss: 1.3156 - val_loss: 1.1617\n",
      "Epoch 4/100\n",
      "13568/14154 [===========================>..] - ETA: 0s - loss: 1.2666\n",
      "Epoch 00004: val_loss improved from 1.16167 to 1.14178, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 1s 99us/sample - loss: 1.2640 - val_loss: 1.1418\n",
      "Epoch 5/100\n",
      "13568/14154 [===========================>..] - ETA: 0s - loss: 1.2017\n",
      "Epoch 00005: val_loss did not improve from 1.14178\n",
      "14154/14154 [==============================] - 1s 88us/sample - loss: 1.2025 - val_loss: 1.1486\n",
      "Epoch 6/100\n",
      "13792/14154 [============================>.] - ETA: 0s - loss: 1.1728\n",
      "Epoch 00006: val_loss improved from 1.14178 to 1.12371, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 1s 97us/sample - loss: 1.1704 - val_loss: 1.1237\n",
      "Epoch 7/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "13696/14154 [============================>.] - ETA: 0s - loss: 1.1421\n",
      "Epoch 00007: val_loss did not improve from 1.12371\n",
      "14154/14154 [==============================] - 1s 87us/sample - loss: 1.1388 - val_loss: 1.1314\n",
      "Epoch 8/100\n",
      "13920/14154 [============================>.] - ETA: 0s - loss: 1.1286\n",
      "Epoch 00008: val_loss improved from 1.12371 to 1.11520, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 1s 98us/sample - loss: 1.1278 - val_loss: 1.1152\n",
      "Epoch 9/100\n",
      "13920/14154 [============================>.] - ETA: 0s - loss: 1.1075\n",
      "Epoch 00009: val_loss improved from 1.11520 to 1.11500, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 1s 98us/sample - loss: 1.1084 - val_loss: 1.1150\n",
      "Epoch 10/100\n",
      "13696/14154 [============================>.] - ETA: 0s - loss: 1.0958\n",
      "Epoch 00010: val_loss did not improve from 1.11500\n",
      "14154/14154 [==============================] - 1s 90us/sample - loss: 1.0973 - val_loss: 1.1189\n",
      "Epoch 11/100\n",
      "13856/14154 [============================>.] - ETA: 0s - loss: 1.0850\n",
      "Epoch 00011: val_loss did not improve from 1.11500\n",
      "14154/14154 [==============================] - 1s 85us/sample - loss: 1.0811 - val_loss: 1.1249\n",
      "Epoch 12/100\n",
      "13792/14154 [============================>.] - ETA: 0s - loss: 1.0632\n",
      "Epoch 00012: val_loss improved from 1.11500 to 1.10706, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 1s 95us/sample - loss: 1.0625 - val_loss: 1.1071\n",
      "Epoch 13/100\n",
      "13472/14154 [===========================>..] - ETA: 0s - loss: 1.0391\n",
      "Epoch 00013: val_loss improved from 1.10706 to 1.10617, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 1s 96us/sample - loss: 1.0413 - val_loss: 1.1062\n",
      "Epoch 14/100\n",
      "13728/14154 [============================>.] - ETA: 0s - loss: 1.0263\n",
      "Epoch 00014: val_loss improved from 1.10617 to 1.09419, saving model to nn_model.w8\n",
      "14154/14154 [==============================] - 1s 103us/sample - loss: 1.0301 - val_loss: 1.0942\n",
      "Epoch 15/100\n",
      "13568/14154 [===========================>..] - ETA: 0s - loss: 1.0174\n",
      "Epoch 00015: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 87us/sample - loss: 1.0189 - val_loss: 1.1066\n",
      "Epoch 16/100\n",
      "13824/14154 [============================>.] - ETA: 0s - loss: 1.0153\n",
      "Epoch 00016: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 84us/sample - loss: 1.0151 - val_loss: 1.1007\n",
      "Epoch 17/100\n",
      "13824/14154 [============================>.] - ETA: 0s - loss: 0.9885\n",
      "Epoch 00017: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 84us/sample - loss: 0.9866 - val_loss: 1.1343\n",
      "Epoch 18/100\n",
      "13888/14154 [============================>.] - ETA: 0s - loss: 0.9884\n",
      "Epoch 00018: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 84us/sample - loss: 0.9886 - val_loss: 1.0971\n",
      "Epoch 19/100\n",
      "14080/14154 [============================>.] - ETA: 0s - loss: 0.9874\n",
      "Epoch 00019: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 86us/sample - loss: 0.9877 - val_loss: 1.1155\n",
      "Epoch 20/100\n",
      "13856/14154 [============================>.] - ETA: 0s - loss: 0.9699\n",
      "Epoch 00020: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 85us/sample - loss: 0.9667 - val_loss: 1.1037\n",
      "Epoch 21/100\n",
      "13792/14154 [============================>.] - ETA: 0s - loss: 0.9623\n",
      "Epoch 00021: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 85us/sample - loss: 0.9610 - val_loss: 1.0994\n",
      "Epoch 22/100\n",
      "13760/14154 [============================>.] - ETA: 0s - loss: 0.9475\n",
      "Epoch 00022: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 85us/sample - loss: 0.9531 - val_loss: 1.1015\n",
      "Epoch 23/100\n",
      "13664/14154 [===========================>..] - ETA: 0s - loss: 0.9414\n",
      "Epoch 00023: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 86us/sample - loss: 0.9441 - val_loss: 1.1074\n",
      "Epoch 24/100\n",
      "13600/14154 [===========================>..] - ETA: 0s - loss: 0.9362\n",
      "Epoch 00024: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 86us/sample - loss: 0.9373 - val_loss: 1.1097\n",
      "Epoch 25/100\n",
      "13792/14154 [============================>.] - ETA: 0s - loss: 0.9247\n",
      "Epoch 00025: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 98us/sample - loss: 0.9237 - val_loss: 1.0995\n",
      "Epoch 26/100\n",
      "13984/14154 [============================>.] - ETA: 0s - loss: 0.9159\n",
      "Epoch 00026: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 2s 108us/sample - loss: 0.9171 - val_loss: 1.1135\n",
      "Epoch 27/100\n",
      "13824/14154 [============================>.] - ETA: 0s - loss: 0.9066\n",
      "Epoch 00027: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 94us/sample - loss: 0.9092 - val_loss: 1.1092\n",
      "Epoch 28/100\n",
      "13952/14154 [============================>.] - ETA: 0s - loss: 0.9050\n",
      "Epoch 00028: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 99us/sample - loss: 0.9032 - val_loss: 1.1295\n",
      "Epoch 29/100\n",
      "14016/14154 [============================>.] - ETA: 0s - loss: 0.8989\n",
      "Epoch 00029: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 105us/sample - loss: 0.8989 - val_loss: 1.1167\n",
      "Epoch 30/100\n",
      "13728/14154 [============================>.] - ETA: 0s - loss: 0.8957\n",
      "Epoch 00030: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 93us/sample - loss: 0.8982 - val_loss: 1.1387\n",
      "Epoch 31/100\n",
      "13920/14154 [============================>.] - ETA: 0s - loss: 0.8798\n",
      "Epoch 00031: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 94us/sample - loss: 0.8782 - val_loss: 1.1249\n",
      "Epoch 32/100\n",
      "13472/14154 [===========================>..] - ETA: 0s - loss: 0.8740\n",
      "Epoch 00032: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 94us/sample - loss: 0.8706 - val_loss: 1.1247\n",
      "Epoch 33/100\n",
      "13888/14154 [============================>.] - ETA: 0s - loss: 0.8613\n",
      "Epoch 00033: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 99us/sample - loss: 0.8616 - val_loss: 1.1603\n",
      "Epoch 34/100\n",
      "14080/14154 [============================>.] - ETA: 0s - loss: 0.8636\n",
      "Epoch 00034: val_loss did not improve from 1.09419\n",
      "14154/14154 [==============================] - 1s 102us/sample - loss: 0.8635 - val_loss: 1.1493\n",
      "------------------------------\n",
      "Partial score of fold 4 is: 0.5380070784335882\n",
      "------------------------------\n",
      "Our oof cohen kappa score is:  0.5703735473976528\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "nn_model = Nn_Model(new_train, new_test, features, categoricals=cat_features)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Ensemble Weighting\n",
    "\n",
    "Ensemble predictions from the 4 models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "weights = {'lbg': 0.25, 'cat': 0.25, 'xgb': 0.25, 'nn': 0.25}\n",
    "\n",
    "preds = (lgb_model.y_pred * weights['lbg']) + cat_model.y_pred * weights['cat'] + (xgb_model.y_pred * weights['xgb']) + (nn_model.y_pred * weights['nn'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The prediction of xgb and lgb are positively correlated. The ideal is they do not correlate significantly, so that ensemble model would extract different information from differnt inner models."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x720 with 6 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "model_preds = {'lgb': lgb_model.y_pred, \n",
    "               'xgb': xgb_model.y_pred,\n",
    "               'catb': cat_model.y_pred, \n",
    "               'nn': nn_model.y_pred}\n",
    "\n",
    "plt.figure(figsize=(15,10))\n",
    "plt_idx = 1\n",
    "for i in range (0, len(model_preds)-1):\n",
    "    i_key = list(model_preds.keys())[i]\n",
    "    for j in range(i+1, len(model_preds)):\n",
    "        j_key = list(model_preds.keys())[j]\n",
    "        plt.subplot(3, 3, plt_idx)\n",
    "        sns.scatterplot(model_preds[i_key], model_preds[j_key])\n",
    "        plt.title(i_key + '-' + j_key)\n",
    "        plt_idx += 1\n",
    "plt.suptitle('Pairwise predictions of all models')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x216 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(15,3))\n",
    "\n",
    "for idx, model in enumerate(model_preds):\n",
    "    plt.subplot(1, len(model_preds), idx+1)\n",
    "    sns.distplot(model_preds[model])\n",
    "    plt.title(model)\n",
    "\n",
    "plt.suptitle('Dist of predictions')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Processing Submission"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "dist = Counter(new_train['accuracy_group'])\n",
    "for k in dist:\n",
    "    dist[k] /= len(new_train)\n",
    "\n",
    "acum = 0\n",
    "bound = {}\n",
    "for i in range(3):\n",
    "    acum += dist[i]\n",
    "    bound[i] = np.percentile(preds, acum * 100) # Returns the q-th percentile (which is acum*100 percentile) of the predictions\n",
    "\n",
    "\n",
    "def classify(x):\n",
    "    if x <= bound[0]:\n",
    "        return 0\n",
    "    elif x <= bound[1]:\n",
    "        return 1\n",
    "    elif x <= bound[2]:\n",
    "        return 2\n",
    "    else:\n",
    "        return 3\n",
    "\n",
    "    \n",
    "final_preds = np.array(list(map(classify, preds)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "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>installation_id</th>\n",
       "      <th>accuracy_group</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>00abaee7</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>01242218</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>017c5718</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>01a44906</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>01bc6cb6</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  installation_id  accuracy_group\n",
       "0        00abaee7               3\n",
       "1        01242218               3\n",
       "2        017c5718               3\n",
       "3        01a44906               3\n",
       "4        01bc6cb6               3"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "submission['accuracy_group'] = final_preds.astype(int)\n",
    "submission.to_csv('submission.csv', index=None)\n",
    "submission.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3    500\n",
       "0    239\n",
       "1    137\n",
       "2    124\n",
       "Name: accuracy_group, dtype: int64"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "submission['accuracy_group'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_solution_string(sub: pd.DataFrame) -> str:\n",
    "    \"\"\"\n",
    "    Run this in your notebook with your test submission as input\n",
    "    to get the output string and copy/paste into the cell below.\n",
    "    \n",
    "    e.g. get_solution_string(my_submission)\n",
    "    \"\"\"\n",
    "    return (sub.installation_id + sub.accuracy_group.map(str)).str.cat()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# get_solution_string(submission)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Other things that not working\n",
    "- https://www.kaggle.com/c/data-science-bowl-2019/discussion/124659\n",
    "- https://www.kaggle.com/c/data-science-bowl-2019/discussion/125792"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## References\n",
    "This work is based on [this baseline solution](https://www.kaggle.com/braquino/convert-to-regression). Big kudos to the author for sharing this solution as we can learn so much from him."
   ]
  }
 ],
 "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": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}