{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "28740129",
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "import os\n",
    "import warnings\n",
    "os.environ['OPENBLAS_NUM_THREADS'] = '1'\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "b907cd02",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import time\n",
    "import pyarrow.parquet as pq\n",
    "import scipy\n",
    "import implicit\n",
    "import bisect\n",
    "import sklearn.metrics as m\n",
    "from catboost import CatBoostClassifier, CatBoostRegressor, Pool\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.calibration import calibration_curve, CalibratedClassifierCV"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "e762eacd",
   "metadata": {},
   "outputs": [],
   "source": [
    "LOCAL_DATA_PATH = './context_data/'\n",
    "SPLIT_SEED = 42\n",
    "DATA_FILE = 'competition_data_final_pqt'\n",
    "TARGET_FILE = 'public_train.pqt'\n",
    "SUBMISSION_FILE = 'submit.pqt'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "id": "8c1af0db",
   "metadata": {},
   "outputs": [],
   "source": [
    "id_to_submit = pq.read_table(f'{LOCAL_DATA_PATH}/{SUBMISSION_FILE}').to_pandas()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "335226b7",
   "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 tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>field</th>\n",
       "      <th>type</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>region_name</td>\n",
       "      <td>string</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>city_name</td>\n",
       "      <td>string</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>cpe_manufacturer_name</td>\n",
       "      <td>string</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>cpe_model_name</td>\n",
       "      <td>string</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>url_host</td>\n",
       "      <td>string</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>cpe_type_cd</td>\n",
       "      <td>string</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>cpe_model_os_type</td>\n",
       "      <td>string</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>price</td>\n",
       "      <td>double</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>date</td>\n",
       "      <td>date32[day]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>part_of_day</td>\n",
       "      <td>string</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>request_cnt</td>\n",
       "      <td>int64</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>user_id</td>\n",
       "      <td>int64</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    field         type\n",
       "0             region_name       string\n",
       "1               city_name       string\n",
       "2   cpe_manufacturer_name       string\n",
       "3          cpe_model_name       string\n",
       "4                url_host       string\n",
       "5             cpe_type_cd       string\n",
       "6       cpe_model_os_type       string\n",
       "7                   price       double\n",
       "8                    date  date32[day]\n",
       "9             part_of_day       string\n",
       "10            request_cnt        int64\n",
       "11                user_id        int64"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data = pq.read_table(f'{LOCAL_DATA_PATH}/{DATA_FILE}')\n",
    "pd.DataFrame([(z.name, z.type) for z in data.schema], columns = [['field', 'type']])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dc1e0c72",
   "metadata": {},
   "source": [
    "Регион   \n",
    ", населенный пункт   \n",
    ", производиель устройства   \n",
    ", модель устроства   \n",
    ", домен, с которого пришел рекламный запрос   \n",
    ", тип устройства (смартфон или что-то другое)   \n",
    ", операционка на устройстве   \n",
    ", оценка цены устройства   \n",
    ", дата   \n",
    ", время дня (утро, вечер ...)   \n",
    ", число запросов   \n",
    ", id пользователя   "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "bb8abea4",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "smartphone    322781599\n",
       "tablet            53768\n",
       "plain             36116\n",
       "phablet           27952\n",
       "Name: cpe_type_cd, dtype: int64"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.select(['cpe_type_cd']).to_pandas()['cpe_type_cd'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "3ff50f46",
   "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 tr th {\n",
       "        text-align: left;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr>\n",
       "      <th></th>\n",
       "      <th>field</th>\n",
       "      <th>type</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>age</td>\n",
       "      <td>double</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>is_male</td>\n",
       "      <td>string</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>user_id</td>\n",
       "      <td>int64</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>__index_level_0__</td>\n",
       "      <td>int64</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               field    type\n",
       "0                age  double\n",
       "1            is_male  string\n",
       "2            user_id   int64\n",
       "3  __index_level_0__   int64"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "targets = pq.read_table(f'{LOCAL_DATA_PATH}/{TARGET_FILE}')\n",
    "pd.DataFrame([(z.name, z.type) for z in targets.schema], columns = [['field', 'type']])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "f6f543bf",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 34.7 s, sys: 6.29 s, total: 41 s\n",
      "Wall time: 40.9 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "data_agg = data.select(['user_id', 'url_host', 'request_cnt']).\\\n",
    "    group_by(['user_id', 'url_host']).aggregate([('request_cnt', \"sum\")])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "57c55747",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "199683 urls\n",
      "415317 users\n"
     ]
    }
   ],
   "source": [
    "url_set = set(data_agg.select(['url_host']).to_pandas()['url_host'])\n",
    "print(f'{len(url_set)} urls')\n",
    "url_dict = {url: idurl for url, idurl in zip(url_set, range(len(url_set)))}\n",
    "usr_set = set(data_agg.select(['user_id']).to_pandas()['user_id'])\n",
    "print(f'{len(usr_set)} users')\n",
    "usr_dict = {usr: user_id for usr, user_id in zip(usr_set, range(len(usr_set)))}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "5e227779",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 8.27 s, sys: 1.84 s, total: 10.1 s\n",
      "Wall time: 9.99 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "values = np.array(data_agg.select(['request_cnt_sum']).to_pandas()['request_cnt_sum'])\n",
    "rows = np.array(data_agg.select(['user_id']).to_pandas()['user_id'].map(usr_dict))\n",
    "cols = np.array(data_agg.select(['url_host']).to_pandas()['url_host'].map(url_dict))\n",
    "mat = scipy.sparse.coo_matrix((values, (rows, cols)), shape=(rows.max() + 1, cols.max() + 1))\n",
    "als = implicit.approximate_als.FaissAlternatingLeastSquares(factors = 50, iterations = 30, use_gpu = False, \\\n",
    "       calculate_training_loss = False, regularization = 0.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "8cf9c775",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 30/30 [00:26<00:00,  1.13it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 57min 12s, sys: 3min 32s, total: 1h 45s\n",
      "Wall time: 32.6 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "als.fit(mat)\n",
    "u_factors = als.model.user_factors \n",
    "d_factors = als.model.item_factors"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ed2944b0",
   "metadata": {},
   "source": [
    "## Получим оценку по полу"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "id": "c2f7f21d",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 1.11 s, sys: 259 ms, total: 1.37 s\n",
      "Wall time: 1.36 s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "1    135331\n",
       "0    128994\n",
       "Name: is_male, dtype: int64"
      ]
     },
     "execution_count": 102,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "inv_usr_map = {v: k for k, v in usr_dict.items()}\n",
    "usr_emb = pd.DataFrame(u_factors)\n",
    "usr_emb['user_id'] = usr_emb.index.map(inv_usr_map)\n",
    "usr_targets = targets.to_pandas()\n",
    "df = usr_targets.merge(usr_emb, how = 'inner', on = ['user_id'])\n",
    "df = df[df['is_male'] != 'NA']\n",
    "df = df.dropna()\n",
    "df['is_male'] = df['is_male'].map(int)\n",
    "df['is_male'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "id": "1419fb8a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GINI по полу 0.657\n",
      "CPU times: user 40min 43s, sys: 1min 24s, total: 42min 7s\n",
      "Wall time: 23.4 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "x_train, x_test, y_train, y_test = train_test_split(\\\n",
    "    df.drop(['user_id', 'age', 'is_male'], axis = 1), df['is_male'], test_size = 0.33, random_state = SPLIT_SEED)\n",
    "clf = CatBoostClassifier()\n",
    "clf.fit(x_train, y_train, verbose = False)\n",
    "print(f'GINI по полу {2 * m.roc_auc_score(y_test, clf.predict_proba(x_test)[:,1]) - 1:2.3f}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "id": "8ee3c4c0",
   "metadata": {},
   "outputs": [],
   "source": [
    "clf.fit(df.drop(['user_id', 'age', 'is_male'], axis = 1), df['is_male'], verbose = False)\n",
    "id_to_submit['is_male'] = clf.predict_proba(id_to_submit.merge(usr_emb, how = 'inner', on = ['user_id']))[:,1]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a3980236",
   "metadata": {},
   "source": [
    "## Получим оценку по возрасту"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "id": "4a4f4726",
   "metadata": {},
   "outputs": [],
   "source": [
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "import plotly.express as px\n",
    "\n",
    "%matplotlib inline\n",
    "sns.set_style('darkgrid')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "id": "0f97ec5d",
   "metadata": {},
   "outputs": [],
   "source": [
    "def age_bucket(x):\n",
    "    return bisect.bisect_left([18,25,35,45,55,65], x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "id": "5d38c185",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot: xlabel='age', ylabel='Count'>"
      ]
     },
     "execution_count": 107,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df = usr_targets.merge(usr_emb, how = 'inner', on = ['user_id'])\n",
    "df = df[df['age'] != 'NA']\n",
    "df = df.dropna()\n",
    "df['age'] = df['age'].map(age_bucket)\n",
    "sns.histplot(df['age'], bins = 7)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "id": "e51e16fb",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "         <18       0.00      0.00      0.00       349\n",
      "       18-25       0.50      0.29      0.36     10663\n",
      "       25-34       0.47      0.62      0.53     28815\n",
      "       35-44       0.39      0.51      0.44     25791\n",
      "       45-54       0.36      0.18      0.24     13931\n",
      "       55-65       0.40      0.19      0.26      7688\n",
      "         65+       0.37      0.02      0.03      1849\n",
      "\n",
      "    accuracy                           0.43     89086\n",
      "   macro avg       0.35      0.26      0.27     89086\n",
      "weighted avg       0.42      0.43      0.40     89086\n",
      "\n"
     ]
    }
   ],
   "source": [
    "x_train, x_test, y_train, y_test = train_test_split(\\\n",
    "    df.drop(['user_id', 'age', 'is_male'], axis = 1), df['age'], test_size = 0.33, random_state = SPLIT_SEED)\n",
    "\n",
    "clf = CatBoostClassifier()\n",
    "clf.fit(x_train, y_train, verbose = False)\n",
    "print(m.classification_report(y_test, clf.predict(x_test), \\\n",
    "                            target_names = ['<18', '18-25','25-34', '35-44', '45-54', '55-65', '65+']))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "id": "2c73661d",
   "metadata": {},
   "outputs": [],
   "source": [
    "clf.fit(df.drop(['user_id', 'age', 'is_male'], axis = 1), df['age'], verbose = False)\n",
    "id_to_submit['age'] = clf.predict(id_to_submit[['user_id']].merge(usr_emb, how = 'inner', on = ['user_id']))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f55aa8ec",
   "metadata": {},
   "source": [
    "## Сабмит"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "id": "af2b6647",
   "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>user_id</th>\n",
       "      <th>is_male</th>\n",
       "      <th>age</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6</td>\n",
       "      <td>0.330467</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>11</td>\n",
       "      <td>0.725477</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>19</td>\n",
       "      <td>0.240190</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>27</td>\n",
       "      <td>0.536798</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>32</td>\n",
       "      <td>0.471325</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   user_id   is_male  age\n",
       "0        6  0.330467    2\n",
       "1       11  0.725477    5\n",
       "2       19  0.240190    1\n",
       "3       27  0.536798    2\n",
       "4       32  0.471325    3"
      ]
     },
     "execution_count": 110,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "id_to_submit.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "id": "60024ea9",
   "metadata": {},
   "outputs": [],
   "source": [
    "id_to_submit.to_csv(f'{LOCAL_DATA_PATH}/submission.csv', index = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "id": "a7298b7b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "user_id,is_male,age\r\n",
      "6,0.330467150589351,2\r\n",
      "11,0.7254769930049977,5\r\n",
      "19,0.24019020466489424,1\r\n",
      "27,0.5367979653267113,2\r\n",
      "32,0.4713251899911531,3\r\n",
      "37,0.2810748555581949,2\r\n",
      "43,0.6659790932425269,2\r\n",
      "44,0.9189155263784968,1\r\n",
      "46,0.5166941298660128,3\r\n"
     ]
    }
   ],
   "source": [
    "! head $LOCAL_DATA_PATH/submission.csv"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4d6a7a85",
   "metadata": {},
   "source": [
    "# Скор на лидерборде"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 151,
   "id": "7164aa61",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.4715992278434493"
      ]
     },
     "execution_count": 151,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "context_scorer(submission, answers)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "horovod2",
   "language": "python",
   "name": "horovod2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}