week7_zaharov.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Идентификация пользователей по посещенным веб-страницам"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "**Примечание для проверяющего.** Информация по всем упомянутым разделам в 7 недели курса **содержится** в данном ноутбуке, но некоторые разделы переименованы и объединены для удобства восприятия."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1. Цели и задачи проекта"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "В данном проекте создаётся классификатор, предсказывающий пользователя по данным о веб-страницах, которые он посетил. Пользователи отличаются между собой по набору сайтов, по датам и времени посещения сайтов, по интервалам нахождения на конкретных сайтах и др.\n",
    "\n",
    "Мы будем использовать продвинутые методы, отказавшись от ручного подхода: делать предсказания эмперически (т.е. на основе опыта человека). Во-первых, эмперические предсказания требуют человеческих затрат времени, во-вторых с помощью предсказательного моделирования мы можем обнаружить интересные закономерности, которые человек не всегда может описать. В-третьих, мы не привязываемся к конкретному человеку, который должен следить за трафиком сети.\n",
    "\n",
    "Так, данный классификатор можно легко приспособить для любых типов пользователей, в том числе для идентификации пользователей, которые тратят слишком много времени на нежелательных сайтах. Просмотр посещенных сайтов может осуществлять робот, что и надёжнее для работодателя, и этичнее по отношению к сотрудникам. Не отвлекаясь напрасно на списки посещенных сайтов, не тратя время, можно получать уведомление лишь в том случае, если Ваш сотрудник явно нарушает рабочий режим."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2. Описание исходных данных"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Данные представляют собой последовательные записи информации о каждом посещённом сайте в виде:\n",
    "\n",
    "$(site_1, time_1, user_1), (site_2, time_2, user_1), ...(site_k, time_k, user_m)$\n",
    "\n",
    "Т.е. мы имеем тройки, где $site_i$ — помещённый пользователем сайт, $time_i$ — время посещения сайта $site_k$, а $user_j$ — обозначение пользователя, который посетил $site_i$ во время $time_i$.\n",
    "\n",
    "Эти данные мы используем для обучения классификатора. При этом в случае нежелательных сайтов можно подготовить разметку пользователя $user_i$ вручную: зная, например, кого-нибудь в организации, кто практически всё своё время проводит на развлекательных сайтах. Лучше, конечно, найти нескольких таких пользователей. И особенно хорошо этот подход должен работать, если у организации имеются свои сайты, над усовершенствованием которых трудятся сотрудники. В этом случае классификатор будет срабатывать, если пользователь отклонился от основной работы. \n",
    "\n",
    "Так, данные о нежелательных сайтах можно, например, собирать по тем дням, когда начальник отсутствовал на рабочем месте. Многие сотрудники при этом работают более расслабленно.\n",
    "\n",
    "В рабочем режиме классификатора у нас не будет информации о пользователе $user_i$ (это и будет работа классификатора — определить пользователя).\n",
    "\n",
    "Кроме того, данные поделены на сессии. Т.е. на предсказании известно, что конкретная последовательность из нескольких пар $site_i, time_i$ принадлежит только одному пользователю, правда, неизвестно кому конкретно. Иначе, если бы сессии пользователей были перемешаны по пользователям — в предсказаниях возникла бы трудность даже у человека."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3. Код контроля задач, используемый в ноутбуке"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Было решено использовать собственное **средство изоляции ячеек.** Реализация в виде класса *Task* ниже.\n",
    "\n",
    "Внутри каждой ячейки выполняются изолированная от глобальных переменных задача. Внешние переменные, используемые в задаче, не являются глобальными, а приходят в качестве формальных параметров в функцию задачи от предыдущей задачи. Это решает следующие проблемы:\n",
    "\n",
    "1. Внешние глобальные переменные недоступны внутри задачи (недоступность поддержана). Таким образом, можно не придумывать длинные имена для переменных, чтобы они не совпадали с уже использованными глобальными переменными.\n",
    "2. Мы можем быть уверены, что никакая переменная внутри задачи не была случайно инициализирована где-то ещё, в другой ячейке. Это помогает совершать меньше ошибок и экономит время поиска бага — в случае неожиданных результатов мы можем не пролистывать весь код в поисках возможной неправильной инициализации переменной, а можем сосредоточиться на конкретной ячейке, будучи уверенными, что проблема в текущем небольшом фрагменте кода.\n",
    "3. Передача переменных между задачами **по типу pipe** удобна тем, что мы можем не беспокоиться о порядке выполнения ячеек, быть уверенными в том, что нужные ячейки выполнены. В противном случае, будет сгенерировано исключение, указывающее, что недостаёт данных после выполнения какой-то задачи — в этом случае можно сделать вывод, что требуемая предыдущая задача не была выполнена. В случае подобной ошибки значение никаких переменных не будет нарушено и можно просто выполнить недостающую задачу.\n",
    "\n",
    "**Примечание.** Для указания, результат какой задачи будет использовать текущая задача, мы используем декоратор **@use_result().** Если никакого результата не требуется, т.к. задача сама начинает цепочку передачи данных, указанный декоратор использовать не нужно."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from __future__ import division, print_function\n",
    "# отключим всякие предупреждения Anaconda\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "import os\n",
    "import sys\n",
    "import traceback\n",
    "import pickle\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import scipy.sparse as sparse\n",
    "from scipy.sparse import csr_matrix, hstack as sparse_hstack\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.linear_model import SGDClassifier\n",
    "from sklearn.metrics import roc_auc_score\n",
    "from IPython.core.display import display, HTML\n",
    "\n",
    "import re\n",
    "import types\n",
    "import inspect\n",
    "from functools import partial\n",
    "\n",
    "from sklearn.model_selection import train_test_split, cross_val_score, GridSearchCV\n",
    "from sklearn.feature_selection import SelectFromModel\n",
    "from sklearn.linear_model import SGDClassifier\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.metrics import roc_auc_score\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "import pylab\n",
    "\n",
    "from MulticoreTSNE import MulticoreTSNE as TSNE\n",
    "\n",
    "import seaborn as sns\n",
    "\n",
    "from collections import Counter, OrderedDict\n",
    "from datetime import datetime\n",
    "\n",
    "import traceback\n",
    "\n",
    "from sklearn.model_selection import validation_curve, learning_curve\n",
    "\n",
    "\n",
    "class ResultDuplicateNextInfoKeyError(Exception): pass\n",
    "class ResultNotFoundError(Exception): pass\n",
    "\n",
    "\n",
    "class Result(object):\n",
    "    def __init__(self, task_result=None, next_info=None):\n",
    "        self.task_result = task_result\n",
    "        self.next_info = next_info\n",
    "\n",
    "\n",
    "    def __add__(self, other_result):\n",
    "        summary_next_info = {}\n",
    "\n",
    "        for key, val in self.next_info.iteritems():\n",
    "            summary_next_info[key] = val\n",
    "\n",
    "        for key, val in other_result.next_info.iteritems():\n",
    "            if key in summary_next_info:\n",
    "                raise ResultDuplicateNextInfoKeyError(\n",
    "                    'There is a duplicate key %s in next_info when adding 2 results.' % repr(key))\n",
    "\n",
    "            summary_next_info[key] = val\n",
    "\n",
    "        return Result(next_info=summary_next_info)\n",
    "\n",
    "\n",
    "\n",
    "class Task(object):\n",
    "    _TASK_PREFIX = 'task_'\n",
    "    _RE_NO_GLOBALS_REMOVE = '__\\w+__'\n",
    "    _LOCAL_GLOBALS = ['Result']\n",
    "    _DECORATED_FUNC_NAME = '__decorated_func'\n",
    "\n",
    "\n",
    "    def __init__(self):\n",
    "        self._results = {}\n",
    "\n",
    "\n",
    "    def _remove_globals(self, calling_globals):\n",
    "        self._old_globals = {}\n",
    "        for key, val in calling_globals.iteritems():\n",
    "            if    re.match(self._RE_NO_GLOBALS_REMOVE, key) is None and key not in self._LOCAL_GLOBALS \\\n",
    "                    and not isinstance(val, types.ModuleType) and not callable(val):\n",
    "                self._old_globals.update({key: val})\n",
    "\n",
    "        for key in self._old_globals.keys():\n",
    "            del calling_globals[key]\n",
    "\n",
    "\n",
    "    def _restore_globals(self, calling_globals):\n",
    "        for key in self._old_globals.iterkeys():\n",
    "            calling_globals[key] = self._old_globals[key]\n",
    "\n",
    "        self._old_globals = {}\n",
    "\n",
    "\n",
    "    def call(self, name, use_result=None):\n",
    "        calling_globals = globals()\n",
    "\n",
    "        task_func = calling_globals[self._TASK_PREFIX + name]\n",
    "\n",
    "        self._remove_globals(calling_globals)\n",
    "\n",
    "        try:\n",
    "            if task_func.__name__ == self._DECORATED_FUNC_NAME:\n",
    "                result = task_func(name, self._results, use_result)\n",
    "            else:\n",
    "                next_info = use_result.next_info if use_result is not None else {}\n",
    "                result = task_func(**next_info)\n",
    "        except:\n",
    "            print('There is an exception on executing task \\'%s\\': %s' % (name, sys.exc_info()[0]))\n",
    "            print(traceback.format_exc())\n",
    "        else:            \n",
    "            if result is not None:\n",
    "                if name in self._results:\n",
    "                    print('Warning! The result of task \\'%s\\' is rewrited.' % name)\n",
    "\n",
    "                self._results[name] = result\n",
    "\n",
    "            if result is not None:\n",
    "                if result.task_result is not None:\n",
    "                    if isinstance(result.task_result, list) or isinstance(result.task_result, tuple):\n",
    "                        task_result = [str(x) for x in result.task_result]\n",
    "                        task_result = ' '.join(task_result)\n",
    "                    else:\n",
    "                        task_result = result.task_result\n",
    "\n",
    "\n",
    "                    print('Result of task %s:\\n%s' % (name, repr(task_result)))\n",
    "        finally:\n",
    "            self._restore_globals(calling_globals)\n",
    "\n",
    "\n",
    "def use_result(*result_keys):\n",
    "    def func_wrapper(task_func):\n",
    "        def __decorated_func(task_name, results_obj, use_result):\n",
    "            func_kwargs = {}\n",
    "\n",
    "            for result_key in result_keys:\n",
    "                try:\n",
    "                    result_obj = results_obj[result_key]\n",
    "                except KeyError:\n",
    "                    raise ResultNotFoundError( \\\n",
    "'The result (in use_result) \\'%s\\' for task \\'%s\\' is not found. ' + \\\n",
    "'Probably an input task has not been executed yet.' % \\\n",
    "                                              (result_key, task_name))\n",
    "                if use_result is not None:\n",
    "                    result_obj = result_obj + use_result\n",
    "                func_kwargs.update(result_obj.next_info)\n",
    "\n",
    "            return task_func(**func_kwargs)\n",
    "\n",
    "        return __decorated_func\n",
    "    return func_wrapper\n",
    "\n",
    "task = Task()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Добавляем классы **FeatureMaker** и **SitesData2** (вторая версия класса **SitesData,** использованного в предыдущих заданиях), которые синтезируют новые признаки на основе имеющихся данных."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class FeatureMaker(object):\n",
    "    _DT_FORMAT = '%Y-%m-%d %H:%M:%S'\n",
    "    _FACEBOOK_SITENAME = 'facebook.com'\n",
    "    _WWW_PREFIX = 'www.'\n",
    "    _DOT = '.'\n",
    "\n",
    "    NUM_TOP_SITES_INDICATORS = 30\n",
    "    SMALL_TIME_DIFFS_INTERVALS = [40, 20, 10, 5, 3, 1]    \n",
    "\n",
    "    \"Делаем недостающие признаки.\"\n",
    "    def __init__(self, sites_by_freq, session_length, use_new_features):        \n",
    "        self._sites_by_freq = sites_by_freq\n",
    "        self._session_length = session_length\n",
    "        self._top_sites = sites_by_freq.keys()[:self.NUM_TOP_SITES_INDICATORS]        \n",
    "        self._use_new_features = use_new_features\n",
    "\n",
    "\n",
    "    def _set_borders(self, session_start, end_pos):\n",
    "        self._session_start = session_start\n",
    "        self._end_pos = end_pos\n",
    "\n",
    "\n",
    "    def set_user_data(self, user_data):\n",
    "        self._user_data = user_data\n",
    "\n",
    "\n",
    "    def get_columns(self):\n",
    "        \"Какие у нас есть колонки?\"\n",
    "        raise Exception('Not implemented.')\n",
    "\n",
    "\n",
    "    def _last_pos(self):\n",
    "        return self._end_pos if self._end_pos < self._user_data.shape[0] else self._end_pos - 1\n",
    "\n",
    "\n",
    "    def _get_site_index(self, site_name):\n",
    "        return self._sites_by_freq[site_name]\n",
    "\n",
    "    \n",
    "    @classmethod\n",
    "    def _dt_from_str(cls, dt_str):\n",
    "        return datetime.strptime(dt_str, cls._DT_FORMAT)\n",
    "\n",
    "\n",
    "    def _dt_diff_seconds(self, first_index, second_index):\n",
    "        first_dt = self._dt_from_str(self._user_data.ix[first_index].timestamp)\n",
    "        second_dt = self._dt_from_str(self._user_data.ix[second_index].timestamp)\n",
    "\n",
    "        return int((second_dt - first_dt).total_seconds())\n",
    "\n",
    "\n",
    "    def _num_unique_sites(self):\n",
    "        return len(self._sites_by_freq)\n",
    "\n",
    "\n",
    "    def site(self):\n",
    "        \"Индексы сайтов, на которых побывал пользователь за сессию.\"\n",
    "        site_indexes = []\n",
    "\n",
    "        for cur_pos in xrange(self._session_start, self._end_pos):\n",
    "            one_site_data = self._user_data.ix[cur_pos]\n",
    "            site_index = self._get_site_index(one_site_data.site)\n",
    "\n",
    "            site_indexes.append(site_index)\n",
    "\n",
    "        site_indexes += [0 for x in xrange(self._session_length - len(site_indexes))] # дозаполнение нулями\n",
    "\n",
    "        return site_indexes\n",
    "\n",
    "\n",
    "    def time_diffs(self):\n",
    "        \"\"\"\n",
    "            В задании допущена, на мой взгляд, некоточность. \n",
    "            Интервалов должно быть 10, чтобы не терять информацию о времени \n",
    "            пребывания на последнем сайте в сессии.\n",
    "            Иначе некоторые другие признаки будут сформированы ТАКЖЕ с потерей последнего элемента сессии.\n",
    "            Вторая неточность - предлагается дозаполнить time_diff для несуществующих сайтов нулями.\n",
    "            Но ведь ведь в данных есть случаи, когда время прибывания на сайте действительно заняло\n",
    "            0 секунд! Это значит, что так мы потеряем часть данных.\n",
    "\n",
    "            Поэтому я дозаполнил несуществующие данные np.nan, а не 0, \n",
    "            чтобы модель была ближе к действительности.\n",
    "            Что-то делать с этими np.nan можно уже походу \n",
    "            (например, заменять их на -1).\n",
    "        \"\"\"\n",
    "        time_diffs = []\n",
    "\n",
    "        for i in range(self._session_start, self._last_pos()):\n",
    "            time_diffs.append(self._dt_diff_seconds(i, i + 1))\n",
    "        for i in range(self._last_pos(), self._session_start + self._session_length):\n",
    "            if self._use_new_features:\n",
    "                time_diffs.append(-1)\n",
    "            else:\n",
    "                time_diffs.append(0)\n",
    "\n",
    "        if not self._use_new_features:\n",
    "            time_diffs.pop()\n",
    "\n",
    "        self._time_diffs = time_diffs\n",
    "\n",
    "        return time_diffs\n",
    "\n",
    "\n",
    "    def session_timespan(self):\n",
    "        last_pos = self._end_pos - 1\n",
    "        return self._dt_diff_seconds(self._session_start, last_pos)\n",
    "\n",
    "\n",
    "    def unique_sites(self):\n",
    "        return len(self._user_data[self._session_start: self._end_pos].site.unique())\n",
    "\n",
    "\n",
    "    def start_hour(self):\n",
    "        dt = self._dt_from_str(self._user_data.ix[self._session_start].timestamp)\n",
    "        return dt.hour\n",
    "\n",
    "\n",
    "    def day_of_week(self):\n",
    "        dt = self._dt_from_str(self._user_data.ix[self._session_start].timestamp)\n",
    "        return dt.weekday()\n",
    "\n",
    "\n",
    "    def day_month_year(self):\n",
    "        \"Год, месяц, день начала сессии.\"\n",
    "        dt = self._dt_from_str(self._user_data.ix[self._session_start].timestamp)\n",
    "        return (dt.day, dt.month, dt.year)\n",
    "\n",
    "\n",
    "    def through_hour(self):\n",
    "        \"Час начала сессии (с учетом года, месяца и дня).\"\n",
    "        epoch_dt = datetime.fromtimestamp(0)\n",
    "        dt = self._dt_from_str(self._user_data.ix[self._session_start].timestamp)\n",
    "\n",
    "        td = dt - epoch_dt\n",
    "\n",
    "        through_hour = int(td.total_seconds() / 3600.)\n",
    "\n",
    "        return through_hour\n",
    "\n",
    "\n",
    "    def _seconds_of_day_from_dt(self, dt):\n",
    "        SECONDS_IN_MINUTE = 60\n",
    "        SECONDS_IN_HOUR = SECONDS_IN_MINUTE * 60\n",
    "\n",
    "        return dt.hour * SECONDS_IN_HOUR + dt.minute * SECONDS_IN_MINUTE\n",
    "\n",
    "\n",
    "    def second_of_day(self):\n",
    "        \"Среднее Время суток (в секундах, т.е. середина сессии).\"\n",
    "\n",
    "        first_dt = self._dt_from_str(self._user_data.ix[self._session_start].timestamp)\n",
    "        last_dt = self._dt_from_str(self._user_data.ix[self._last_pos()].timestamp)\n",
    "        \n",
    "        first_total_seconds = self._seconds_of_day_from_dt(first_dt)\n",
    "        last_total_seconds = self._seconds_of_day_from_dt(last_dt)\n",
    "\n",
    "        return int((first_total_seconds + last_total_seconds) / 2.)\n",
    "\n",
    "\n",
    "    def avg_seconds_on_top_sites(self):\n",
    "        intervals = []\n",
    "\n",
    "        for i in range(self._last_pos() - self._session_start):\n",
    "            site = self._user_data.ix[i].site\n",
    "\n",
    "            if site in self._top_sites:\n",
    "                intervals.append(self._time_diffs[i])\n",
    "\n",
    "        num_intervals = len(intervals)\n",
    "\n",
    "        if num_intervals > 0:\n",
    "            return float(sum(intervals)) / num_intervals\n",
    "        else:\n",
    "            return 0\n",
    "\n",
    "\n",
    "    def top_sites_ind(self):\n",
    "        indicators = [0] * self.NUM_TOP_SITES_INDICATORS\n",
    "\n",
    "        for i in range(self._session_start, self._end_pos):\n",
    "            site = self._user_data.ix[i].site\n",
    "\n",
    "            try:\n",
    "                site_index = self._top_sites.index(site)\n",
    "            except ValueError:\n",
    "                pass\n",
    "            else:\n",
    "                indicators[site_index] = 1\n",
    "\n",
    "        return indicators\n",
    "\n",
    "\n",
    "    def facebook_num(self):\n",
    "        num = 0\n",
    "\n",
    "        for i in range(self._session_start, self._end_pos):\n",
    "            site = self._user_data.ix[i].site\n",
    "            if site.endswith(self._FACEBOOK_SITENAME):\n",
    "                num += 1\n",
    "\n",
    "        return num\n",
    "\n",
    "\n",
    "    def www_sites(self):\n",
    "        num = 0\n",
    "\n",
    "        for i in range(self._session_start, self._end_pos):\n",
    "            site = self._user_data.ix[i].site\n",
    "\n",
    "            if site.startswith(self._WWW_PREFIX):\n",
    "                num += 1\n",
    "\n",
    "        return num\n",
    "\n",
    "\n",
    "    def small_time_diffs(self):\n",
    "        \"Интервалы задаются в секундах.\"\n",
    "\n",
    "        nums = []\n",
    "\n",
    "        for i_interval in range(len(self.SMALL_TIME_DIFFS_INTERVALS)):\n",
    "            nums.append(0)\n",
    "\n",
    "        for i in range(self._last_pos() - self._session_start):\n",
    "            time_diff = self._time_diffs[i]\n",
    "\n",
    "            for i_interval, interval in enumerate(self.SMALL_TIME_DIFFS_INTERVALS):\n",
    "                # если time_diff определён\n",
    "                if not np.isnan(time_diff): \n",
    "                    if time_diff < interval:\n",
    "                        nums[i_interval] += 1\n",
    "\n",
    "        return nums\n",
    "\n",
    "\n",
    "    def num_next_changes(self):\n",
    "        num = 0\n",
    "\n",
    "        cur_site = self._user_data.ix[0].site\n",
    "\n",
    "        for i in range(self._session_start + 1, self._end_pos):\n",
    "            next_site = self._user_data.ix[i].site\n",
    "\n",
    "            if next_site != cur_site:\n",
    "                num += 1\n",
    "                cur_site = next_site\n",
    "\n",
    "        return num\n",
    "\n",
    "\n",
    "    def num_next_same(self):\n",
    "        num = 0\n",
    "\n",
    "        cur_site = self._user_data.ix[0].site\n",
    "\n",
    "        for i in range(self._session_start + 1, self._end_pos):\n",
    "            next_site = self._user_data.ix[i].site\n",
    "\n",
    "            if next_site == cur_site:\n",
    "                num += 1\n",
    "                cur_site = None\n",
    "            else:\n",
    "                cur_site = next_site\n",
    "\n",
    "        return num\n",
    "\n",
    "\n",
    "    def num_most_freq_site(self):\n",
    "        sites = [self._user_data.ix[i].site for i in range(self._session_start, self._end_pos)]\n",
    "        return Counter(sites).most_common(1)[0][1]\n",
    "\n",
    "\n",
    "    def _nums_dots(self):\n",
    "        return [self._user_data.ix[i].site.count(self._DOT) for i in range(self._session_start, self._end_pos)]\n",
    "\n",
    "\n",
    "    def avg_num_dots(self):\n",
    "        nums = self._nums_dots()\n",
    "        return float(sum(nums) / len(nums))\n",
    "\n",
    "\n",
    "    def min_num_dots(self):\n",
    "        return min(self._nums_dots())\n",
    "\n",
    "\n",
    "    def max_num_dots(self):\n",
    "        return max(self._nums_dots())\n",
    "\n",
    "\n",
    "    def min_time_diff(self):\n",
    "        time_diffs = filter(lambda x: x != -1, self._time_diffs)\n",
    "        if len(time_diffs) == 0:\n",
    "            time_diffs = [0]\n",
    "\n",
    "        return min(time_diffs)\n",
    "\n",
    "\n",
    "    @classmethod\n",
    "    def _ts_hour(cls, timestamp):\n",
    "        return datetime.strptime(timestamp, '%Y-%m-%d %H:%M:%S').hour\n",
    "\n",
    "\n",
    "    def start_0_6(self):\n",
    "        return 1 if 0 <= self._ts_hour(self._user_data.ix[0].timestamp) < 6 else 0\n",
    "\n",
    "\n",
    "    def start_6_12(self):\n",
    "        return 1 if 6 <= self._ts_hour(self._user_data.ix[0].timestamp) < 12 else 0\n",
    "\n",
    "\n",
    "    def start_12_18(self):\n",
    "        return 1 if 12 <= self._ts_hour(self._user_data.ix[0].timestamp) < 18 else 0\n",
    "\n",
    "\n",
    "    def start_18_24(self):\n",
    "        return 1 if 18 <= self._ts_hour(self._user_data.ix[0].timestamp) < 24 else 0\n",
    "\n",
    "\n",
    "    def _get_new_features_frame_info(self):\n",
    "        result = [\n",
    "            (['day', 'month', 'year'], self.day_month_year),\n",
    "            ('through_hour', self.through_hour),\n",
    "            ('second_of_day', self.second_of_day),\n",
    "            ('avg_seconds_on_top_sites', self.avg_seconds_on_top_sites),\n",
    "            (['top_site_ind%d' % (i + 1) for i in range(self.NUM_TOP_SITES_INDICATORS)], self.top_sites_ind),\n",
    "            ('facebook_num', self.facebook_num),\n",
    "            ('www_sites', self.www_sites),\n",
    "            (['small_time_diffs%d' % i for i in self.SMALL_TIME_DIFFS_INTERVALS], self.small_time_diffs),\n",
    "            ('num_next_changes', self.num_next_changes),\n",
    "            ('num_next_same', self.num_next_same),\n",
    "            ('num_most_freq_site', self.num_most_freq_site),\n",
    "            ('avg_num_dots', self.avg_num_dots),\n",
    "            ('min_num_dots', self.min_num_dots),\n",
    "            ('max_num_dots', self.max_num_dots),\n",
    "            ('min_time_diff', self.min_time_diff),\n",
    "            ('start_0_6', self.start_0_6),\n",
    "            ('start_6_12', self.start_6_12),\n",
    "            ('start_12_18', self.start_12_18),\n",
    "            ('start_18_24', self.start_18_24),\n",
    "        ]\n",
    "\n",
    "        return result\n",
    "\n",
    "\n",
    "    def name_site_columns(self):\n",
    "        return ['site%d' % (i + 1) for i in xrange(self._session_length)]\n",
    "\n",
    "\n",
    "\n",
    "    def _get_frame_info(self):\n",
    "        end_timediff = self._session_length + 1 if self._use_new_features else self._session_length\n",
    "\n",
    "        num_unique_sites = self._num_unique_sites()\n",
    "\n",
    "        result = [\n",
    "            (['time_diff' + str(i) for i in xrange(1, end_timediff)], self.time_diffs),\n",
    "            ('session_timespan', self.session_timespan),\n",
    "            ('#unique_sites', self.unique_sites),\n",
    "            ('start_hour', self.start_hour),\n",
    "            ('day_of_week', self.day_of_week)\n",
    "        ]\n",
    "\n",
    "        if self._use_new_features:\n",
    "            result += self._get_new_features_frame_info()\n",
    "\n",
    "        return result\n",
    "\n",
    "\n",
    "    def _get_columns(self, get_func):\n",
    "        cols = []\n",
    "\n",
    "        for col_info, _ in get_func():\n",
    "            if isinstance(col_info, basestring):\n",
    "                cols.append(col_info)\n",
    "            else:\n",
    "                cols += list(col_info)\n",
    "\n",
    "        return cols\n",
    "\n",
    "\n",
    "    def get_columns(self):\n",
    "        return self._get_columns(self._get_frame_info)\n",
    "\n",
    "\n",
    "    def get_feature_row(self, session_start, end_pos):\n",
    "        self._set_borders(session_start, end_pos)\n",
    "\n",
    "        row = []\n",
    "\n",
    "        for key, feature_func in self._get_frame_info():\n",
    "            if callable(feature_func):\n",
    "                feature_val = feature_func()            \n",
    "            else:\n",
    "                feature_val = feature_func\n",
    "\n",
    "            if isinstance(feature_val, (list, tuple)):\n",
    "                len_key = len(key)\n",
    "                len_feature_val = len(feature_val)\n",
    "\n",
    "                if len_key != len_feature_val:\n",
    "                    raise Exception( \\\n",
    "                        'Wrong number feature values generated ' + \n",
    "                        'by feature maker method (len=%d) for columns %s (len must be %d)' \\\n",
    "                        % (len_feature_val, repr(key), len_key))\n",
    "\n",
    "                row += list(feature_val)\n",
    "            else:\n",
    "                row.append(feature_val)\n",
    "        \n",
    "        return row\n",
    "\n",
    "\n",
    "class SitesData2(object):\n",
    "    _SESSION_LENGTH = 10\n",
    "\n",
    "    def __init__(self, path_to_train_csv, path_to_test_csv, path_to_site_dic):\n",
    "        self._path_to_train_csv = path_to_train_csv\n",
    "        self._path_to_test_csv = path_to_test_csv\n",
    "        self._path_to_site_dic = path_to_site_dic    \n",
    "\n",
    "        self._load_sites_by_freq()    \n",
    "\n",
    "\n",
    "    @classmethod\n",
    "    def _load_data(cls, path_to_csv):\n",
    "        return pd.read_csv(path_to_csv, index_col='session_id')\n",
    "\n",
    "\n",
    "    def _load_sites_by_freq(self):\n",
    "        with open(self._path_to_site_dic, 'rb') as f:\n",
    "            self._sites_by_freq = pickle.load(f)\n",
    "            self._freq_by_sites = {v: k for k, v in self._sites_by_freq.iteritems()}\n",
    "\n",
    "        self._feature_maker = FeatureMaker( \\\n",
    "            sites_by_freq=self._sites_by_freq, \n",
    "            session_length=self._SESSION_LENGTH, \n",
    "            use_new_features=True\n",
    "        )\n",
    "\n",
    "\n",
    "    def _prepare_features(self, frame, stage=None):\n",
    "\n",
    "        result_data = []\n",
    "        targets = []\n",
    "\n",
    "        num_rows = frame.shape[0]\n",
    "\n",
    "        for num_row, row in frame.iterrows():\n",
    "            timestamps = row[['time%d' % i for i in range(1, self._SESSION_LENGTH + 1)]]\n",
    "\n",
    "            # здесь должны быть имена сайтов, а не индексы\n",
    "            sites = row[['site%d' % i for i in range(1, self._SESSION_LENGTH + 1)]]\n",
    "            sites = filter(lambda x: not np.isnan(x), sites)\n",
    "            site_names = [self._freq_by_sites[site] for site in sites]\n",
    "            timestamps = timestamps.values[:len(sites)]\n",
    "\n",
    "            user_data = pd.DataFrame(data=dict(timestamp=timestamps, site=site_names))\n",
    "\n",
    "            self._feature_maker.set_user_data(user_data) # в предыдущей версии разбито по-другому, посмотреть\n",
    "            \n",
    "            feature_row = self._feature_maker.get_feature_row(0, user_data.shape[0])\n",
    "            result_data.append(feature_row)\n",
    "\n",
    "            target = row.target if 'target' in row else None\n",
    "            targets.append(target)\n",
    "\n",
    "\n",
    "        return pd.DataFrame(result_data, columns=self._feature_maker.get_columns()), targets\n",
    "\n",
    "\n",
    "    def prepare_train_features(self):\n",
    "        train_frame = self._load_data(self._path_to_train_csv)\n",
    "        self.train_feature_data, train_targets = self._prepare_features(train_frame, 'prepare_train_features')\n",
    "        self.train_targets = np.array(train_targets)\n",
    "\n",
    "\n",
    "    def prepare_test_features(self):\n",
    "        test_frame = self._load_data(self._path_to_test_csv)\n",
    "        self.test_feature_data, _ = self._prepare_features(test_frame, 'prepare_test_features')\n",
    "\n",
    "\n",
    "    def prepare_site_cnts(self):\n",
    "        train_df = self._load_data(self._path_to_train_csv)\n",
    "        test_df = self._load_data(self._path_to_test_csv)\n",
    "\n",
    "        train_test_df = pd.concat([train_df, test_df])\n",
    "\n",
    "        train_test_df_sites = train_test_df[['site%d' % i for i in range(1, 11)]].fillna(0).astype('int')\n",
    "        train_test_sparse = frame_to_csr(train_test_df_sites.values)\n",
    "\n",
    "        nrows_train_df = train_df.shape[0]\n",
    "\n",
    "        self.train_site_cnts_data = train_test_sparse[:nrows_train_df,:]\n",
    "        self.test_site_cnts_data = train_test_sparse[nrows_train_df:,:]\n",
    "\n",
    "\n",
    "    @classmethod\n",
    "    def _pickle(cls, path, frame):\n",
    "        with open(path, 'wb') as f:\n",
    "            pickle.dump(obj=frame, file=f)\n",
    "\n",
    "\n",
    "    def pickle_train_features(self, path):\n",
    "        self._pickle(path, self.train_feature_data)\n",
    "\n",
    "\n",
    "    def pickle_train_targets(self, path):\n",
    "        self._pickle(path, self.train_targets)        \n",
    "\n",
    "\n",
    "    def pickle_test_features(self, path):\n",
    "        self._pickle(path, self.test_feature_data)\n",
    "\n",
    "\n",
    "    def pickle_train_site_cnts(self, path):\n",
    "        self._pickle(path, self.train_site_cnts_data)\n",
    "\n",
    "\n",
    "    def pickle_test_site_cnts(self, path):\n",
    "        self._pickle(path, self.test_site_cnts_data)\n",
    "\n",
    "\n",
    "    @classmethod\n",
    "    def unpickle(cls, path):\n",
    "        return pd.read_pickle(path)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Дополнительный код (графики, преобразование данных в разреженный вид)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "class Singleton(type):\n",
    "    _instances = {}\n",
    "    def __call__(cls, *args, **kwargs):        \n",
    "        if cls not in cls._instances:\n",
    "            cls._instances[cls] = super(Singleton, cls).__call__(*args, **kwargs)\n",
    "\n",
    "        if len(args) > 0 or len(kwargs) > 0:\n",
    "            raise SingletonError(\n",
    "                'The class <%s> is a singleton and it\\'s object can not be initialized more then once.' \\\n",
    "                % cls.__name__)\n",
    "\n",
    "        return cls._instances[cls]\n",
    "    \n",
    "    \n",
    "%matplotlib inline\n",
    "    \n",
    "def mult_dist(data, label, figsize=None, fontsize=None):\n",
    "    \"Распределение нескольких фич с помощью библиотеки seaborn.\"\n",
    "\n",
    "    params = {}\n",
    "\n",
    "    if figsize is not None:\n",
    "        fig, ax = Plotter().get_plt().subplots(figsize=figsize)\n",
    "        params['ax'] = ax\n",
    "\n",
    "    for col in data.columns:\n",
    "        col_data = data[col]\n",
    "\n",
    "        params['data'] = col_data\n",
    "        \n",
    "        sns.kdeplot(**params)\n",
    "\n",
    "    label_params = {'s': label}\n",
    "    if label is not None:\n",
    "        if fontsize is not None:\n",
    "            label_params['fontsize'] = fontsize\n",
    "\n",
    "        Plotter().get_plt().xlabel(**label_params)\n",
    "\n",
    "    Plotter().get_plt().tight_layout()\n",
    "\n",
    "    Plotter().get_plt().show()\n",
    "    \n",
    "def _plot_feature_distr(features, targets, feature_name):\n",
    "    \"Оценить распределения фичи feature_name при разных target.\"\n",
    "\n",
    "    custom_features = features[feature_name]\n",
    "    data = pd.concat([custom_features, pd.DataFrame(data=dict(target=targets))], axis=1)\n",
    "\n",
    "    target_classes = np.unique(targets)\n",
    "\n",
    "    plot_dict = {}\n",
    "    for target_class in target_classes:\n",
    "        plot_dict.update({'%s (target=%d)' % (feature_name, target_class): \\\n",
    "                          data[data.target==target_class][feature_name]})\n",
    "\n",
    "    plot_data = pd.DataFrame(data=plot_dict)\n",
    "\n",
    "    mult_dist(plot_data, label='Distribution of %s' % feature_name)\n",
    "    \n",
    "    \n",
    "class DisabledCls(object):\n",
    "    \"При вызове любого метода, а также взятии индекса - ничего не делается.\"\n",
    "    def __init__(self, *args, **kwargs):\n",
    "        pass\n",
    "\n",
    "\n",
    "    def __call__(self, *args, **kwargs):\n",
    "        return DisabledCls()\n",
    "\n",
    "\n",
    "    def __getitem__(self, key):\n",
    "        return DisabledCls()\n",
    "\n",
    "\n",
    "    def __getattr__(self, attr):\n",
    "        return DisabledCls()\n",
    "\n",
    "\n",
    "\n",
    "class Plotter(object):\n",
    "    \"\"\"\n",
    "    Вспомогательный класс для отключения графиков (если идёт работа в обычной консоли Python,\n",
    "    а не в ноутбуке).\n",
    "    \"\"\"\n",
    "    __metaclass__ = Singleton\n",
    "\n",
    "    def set_enabled(self, enabled):\n",
    "        self._enabled = enabled\n",
    "\n",
    "\n",
    "    def get_plt(self):\n",
    "        if self._enabled:\n",
    "            return plt\n",
    "        else:\n",
    "            return DisabledCls()\n",
    "        \n",
    "        \n",
    "def frame_to_csr(X):\n",
    "    \"\"\"\n",
    "    Входные данные. 2d массив (матрица), где каждая строка \n",
    "    - сессия с номерами сайтов (а не их частотами-количествами).\n",
    "    Результат. Разреженная матрица частот (количеств), где частота \n",
    "    на пересечении строки (номер сессии) и столбца (индекс сайта).    \n",
    "    \"\"\"\n",
    "\n",
    "    csr_rows = []\n",
    "    csr_cols = []\n",
    "    csr_data = []\n",
    "\n",
    "    for num_row in xrange(X.shape[0]):\n",
    "\n",
    "        row = X[num_row]\n",
    "        num_occurencies = dict(zip(*np.unique(row, return_counts=True)))\n",
    "\n",
    "        for element, num_element_occurencies in num_occurencies.iteritems():\n",
    "            csr_rows.append(num_row)\n",
    "            csr_cols.append(element)\n",
    "            csr_data.append(num_element_occurencies)\n",
    "\n",
    "    csr_X = sparse.csr_matrix((csr_data, (csr_rows, csr_cols)), shape=(X.shape[0], np.max(csr_cols) + 1))    \n",
    "    csr_X = csr_X[:,1:] # первый столбец смысла не имеет, количество повторений 0 - такого индекса нет\n",
    "\n",
    "    return csr_X\n",
    "\n",
    "\n",
    "class HistParams(object):\n",
    "    \"Параметры гистограммы.\"\n",
    "    def __init__(self, data, title, color, xticklabels=None, legend=None):\n",
    "        self.data = data\n",
    "        self.title = title\n",
    "        self.color = color\n",
    "        self.xticklabels = xticklabels\n",
    "        self.legend = legend\n",
    "\n",
    "\n",
    "def color_hist(x_caption, y_caption, *hist_params, **kwargs):\n",
    "    \"Код для рисования гистограмм.\"\n",
    "    _DEFAULT_FIGSIZE = (16, 10)\n",
    "    \n",
    "    MAX_COLS = 2\n",
    "    LEGEND_LOC = 'upper left'\n",
    "    \n",
    "    KEY_ROTATION = 'rotation'\n",
    "    \n",
    "    rotation = kwargs[KEY_ROTATION] if KEY_ROTATION in kwargs else 0\n",
    "\n",
    "    num_hists = len(hist_params)\n",
    "\n",
    "    # автоматически вычисляем параметры строк и колонок для подграфиков\n",
    "    # (с учетом того, что количество колонок не может быть более 4)\n",
    "    ncols = min(num_hists, MAX_COLS)\n",
    "    nrows = int((num_hists - 1) / ncols) + 1\n",
    "\n",
    "    figsize_dx, figsize_dy = _DEFAULT_FIGSIZE\n",
    "\n",
    "    d_coeff = ncols / float(MAX_COLS)\n",
    "\n",
    "    figsize_dx = d_coeff * figsize_dx\n",
    "    figsize_dy = d_coeff * figsize_dy\n",
    "\n",
    "    fig, axs = plt.subplots(nrows=nrows, ncols=ncols, figsize=(figsize_dx, figsize_dy))\n",
    "\n",
    "    for i, hist_param in enumerate(hist_params):\n",
    "        if num_hists > 1:\n",
    "            ax = axs[i] if num_hists > 1 else axs\n",
    "            k = int(i / MAX_COLS)\n",
    "            l = i % MAX_COLS\n",
    "        else:\n",
    "            ax = axs\n",
    "\n",
    "        ax.set_title(hist_param.title)\n",
    "        ax.set_xlabel(x_caption)\n",
    "        ax.set_ylabel(y_caption)\n",
    "        ax.yaxis.grid(True, color='lightgrey')\n",
    "        ax.set_axisbelow(True)\n",
    "\n",
    "        if hist_param.xticklabels is not None:\n",
    "            labels = ax.set_xticklabels(hist_param.xticklabels, rotation=rotation)\n",
    "            num_labels = len(hist_param.xticklabels)\n",
    "            ax.hist(hist_param.data, color=hist_param.color, bins=range(num_labels + 1), align='left', rwidth=0.8)\n",
    "            ax.set_xticks([i for i in range(num_labels)])\n",
    "\n",
    "            if hist_param.legend is not None:\n",
    "                handles, labels = ax.get_legend_handles_labels()\n",
    "                ax.legend([hist_param.legend])\n",
    "        else:\n",
    "            ax.hist(hist_param.data, color=hist_param.color, align='left', rwidth=0.8)\n",
    "    \n",
    "    plt.tight_layout()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4. Предобработка данных"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Исходные данные представлены последовательными записями о каждом посещённом сайте.\n",
    "\n",
    "$(site_1, time_1, user_1), (site_2, time_2, user_1), ...(site_k, time_k, user_m)$\n",
    "\n",
    "Предсказывать пользователя по двойке значений $site_1$, $time_1$ — большого смысла нет, т.к. этих данных слишком мало. Разные пользователи могут посетить один и тот же сайт в одно и то же время.\n",
    "\n",
    "Поэтому будем захватывать некоторый интервал посещений, называемый **сессией.** Будем считать сессией последовательность\n",
    "\n",
    "$(site_i, time_i, user_k), (site_{i+1}, time_{i+1}, user_k), ...(site_n, time_n, user_k)$\n",
    "\n",
    "фиксированной длины. Т.е. будем брать не двойку значений $site_m$, $time_m$ и сопоставлять одному пользователю, а некоторый последовательный набор таких двоек, сопоставленных одному и тому же пользовоателю. Зададим параметр — длину сессии. Это максимальная длина данной последовательности.\n",
    "\n",
    "Каждый сайт переведём в соответствующее число (индекс). Традиционные модели работают только с числами (или с повторяющимися знаковыми последовательностями, воспринимаемыми как категории — не наш случай). Поэтому просмотрим все сайты в выборке и нумеруем их. Один из способов нумерации — пусть сайты, которые встречаются чаще, соответствуют м**е**ньшему индексу. Названия сайтов в модели проигнорируем. Их можно использовать для генерации других признаков, например, признака \"среднее количество точек-разделителей в именах сайтов текущей сессии\". Код, переводящий название сайта в его номер, представлен в методе *top_site_ind()* класса *FeatureMaker.*\n",
    "\n",
    "Каждая сессия должна быть одинаковой длины. Это нужно для того, чтобы воспользоваться традиционными моделями машинного обучения, где количество столбцов с признаками постоянно. Из-за того, что сессия может преждевременно заканчиваться, недополученные значения можно заполнить значениями по умолчанию, например, *нулями.*\n",
    "\n",
    "В сессию можно включить номера сайтов. Код, генерирующий сессию, представлен медотом *_prepare_features()* класса *SitesData2.* \n",
    "\n",
    "Если просто включить номера сайтов в последовательном порядке (первый признак — номер первого сайта в сессии, второй признак — номер второго сайта в сессии и т.д.), то не все модели смогут эффективно обучаться на таком наборе признаков. Причина — признаки перекрываются. Т.е. первый сайт в текущей сессии может случайным образом отличаться от первого сайта в следующей сессии по причине фиксированной длины сессии. Наша сессия — это лишь приближение для обучения, реальный сеанс работы пользователя может быть длиннее. И первый признак может иметь для разных строк (сессий) разный смысл. \n",
    "\n",
    "Поэтому воспользуемся подходом \"мешок слов\" **(bag-of-words)** и будем работать не с последовательными номерами встречающихся сайтов, а используем для информации о сайтах другие признаки: пусть в столбце *k* хранится, сколько раз сайт с индексом *k* встретился в текущей сессии. Так мы получим довольно широкие матрицы признаков с большим количеством нулей, т.к. при небольшой длине сессии (10 сайтов) встретится в ней может не больше 10 разных сайтов, а всего разных сайтов в нашей выборке — десятки тысяч. \n",
    "\n",
    "Для хранения информации о сайтах используем разреженные матрицы (матрицы очень широкие, не поместятся в оперативной памяти). Код генерации признаков \"bag-of-words\" и создания разреженных матриц представлен в функции *frame_to_csr().*\n",
    "\n",
    "Из времен посещения сайтов можно сгенерировать новые признаки. Cгенерируем дополнительные признаки, например\n",
    "\"час начала сессии\", \"временная длина сессии\" и др.\n",
    "\n",
    "Признаки вычисляются на основе одной сессии из 10 сайтов классом *FeatureMaker* (в методах *_get_new_features_frame_info()*, *_get_frame_info()* видны названия признаков и методы, исползуемые для их вычисления). Класс *SitesData* формирует сессию из последотельности двоек $site_m$, $time_m$ и занимается вспомогательными операциями вроде сохранения, считывания данных с диска.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5. Формальный анализ признаков"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Проверим формат исходных данных. Все действия оформляем в виде отдельных задач с целью не допустить случайное изменение переменной другой ячейкой."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Будем работать с данными, разделенными на 2 класса. В начальной выборке есть много пользователей, но задача их разделения по данным посещённых сайтах слишком сложна, чтобы быть выполненной с точностью близкой хотя бы к 80% — так, глядя лишь на эти данные, человек бы регулярно путал одного пользователя с другим. Потому что, когда людей много, всегда найдутся очень похожие по стереотипам поведения. И, т.к. столбцов в данных довольно мало — всего название сайта и время, могут образоваться кластеры пользователей, посещающих одни и те же сайты в близкое время.\n",
    "\n",
    "Поэтому будем работать с данными, где один пользователь отделен от других — попробуем с точностью не ниже 90% отделять целевого пользователя. Здесь и далее — будем называть пользователя именем **Элис.**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>session_id</th>\n",
       "      <th>site1</th>\n",
       "      <th>time1</th>\n",
       "      <th>site2</th>\n",
       "      <th>time2</th>\n",
       "      <th>site3</th>\n",
       "      <th>time3</th>\n",
       "      <th>site4</th>\n",
       "      <th>time4</th>\n",
       "      <th>site5</th>\n",
       "      <th>...</th>\n",
       "      <th>time6</th>\n",
       "      <th>site7</th>\n",
       "      <th>time7</th>\n",
       "      <th>site8</th>\n",
       "      <th>time8</th>\n",
       "      <th>site9</th>\n",
       "      <th>time9</th>\n",
       "      <th>site10</th>\n",
       "      <th>time10</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>718</td>\n",
       "      <td>2014-02-20 10:02:45</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>890</td>\n",
       "      <td>2014-02-22 11:19:50</td>\n",
       "      <td>941.0</td>\n",
       "      <td>2014-02-22 11:19:50</td>\n",
       "      <td>3847.0</td>\n",
       "      <td>2014-02-22 11:19:51</td>\n",
       "      <td>941.0</td>\n",
       "      <td>2014-02-22 11:19:51</td>\n",
       "      <td>942.0</td>\n",
       "      <td>...</td>\n",
       "      <td>2014-02-22 11:19:51</td>\n",
       "      <td>3847.0</td>\n",
       "      <td>2014-02-22 11:19:52</td>\n",
       "      <td>3846.0</td>\n",
       "      <td>2014-02-22 11:19:52</td>\n",
       "      <td>1516.0</td>\n",
       "      <td>2014-02-22 11:20:15</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>2014-02-22 11:20:16</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>14769</td>\n",
       "      <td>2013-12-16 16:40:17</td>\n",
       "      <td>39.0</td>\n",
       "      <td>2013-12-16 16:40:18</td>\n",
       "      <td>14768.0</td>\n",
       "      <td>2013-12-16 16:40:19</td>\n",
       "      <td>14769.0</td>\n",
       "      <td>2013-12-16 16:40:19</td>\n",
       "      <td>37.0</td>\n",
       "      <td>...</td>\n",
       "      <td>2013-12-16 16:40:19</td>\n",
       "      <td>14768.0</td>\n",
       "      <td>2013-12-16 16:40:20</td>\n",
       "      <td>14768.0</td>\n",
       "      <td>2013-12-16 16:40:21</td>\n",
       "      <td>14768.0</td>\n",
       "      <td>2013-12-16 16:40:22</td>\n",
       "      <td>14768.0</td>\n",
       "      <td>2013-12-16 16:40:24</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>782</td>\n",
       "      <td>2014-03-28 10:52:12</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-03-28 10:52:42</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-03-28 10:53:12</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-03-28 10:53:42</td>\n",
       "      <td>782.0</td>\n",
       "      <td>...</td>\n",
       "      <td>2014-03-28 10:54:42</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-03-28 10:55:12</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-03-28 10:55:42</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-03-28 10:56:12</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-03-28 10:56:42</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>22</td>\n",
       "      <td>2014-02-28 10:53:05</td>\n",
       "      <td>177.0</td>\n",
       "      <td>2014-02-28 10:55:22</td>\n",
       "      <td>175.0</td>\n",
       "      <td>2014-02-28 10:55:22</td>\n",
       "      <td>178.0</td>\n",
       "      <td>2014-02-28 10:55:23</td>\n",
       "      <td>177.0</td>\n",
       "      <td>...</td>\n",
       "      <td>2014-02-28 10:55:59</td>\n",
       "      <td>175.0</td>\n",
       "      <td>2014-02-28 10:55:59</td>\n",
       "      <td>177.0</td>\n",
       "      <td>2014-02-28 10:55:59</td>\n",
       "      <td>177.0</td>\n",
       "      <td>2014-02-28 10:57:06</td>\n",
       "      <td>178.0</td>\n",
       "      <td>2014-02-28 10:57:11</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 22 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   session_id  site1                time1  site2                time2  \\\n",
       "0           1    718  2014-02-20 10:02:45    NaN                  NaN   \n",
       "1           2    890  2014-02-22 11:19:50  941.0  2014-02-22 11:19:50   \n",
       "2           3  14769  2013-12-16 16:40:17   39.0  2013-12-16 16:40:18   \n",
       "3           4    782  2014-03-28 10:52:12  782.0  2014-03-28 10:52:42   \n",
       "4           5     22  2014-02-28 10:53:05  177.0  2014-02-28 10:55:22   \n",
       "\n",
       "     site3                time3    site4                time4  site5   ...    \\\n",
       "0      NaN                  NaN      NaN                  NaN    NaN   ...     \n",
       "1   3847.0  2014-02-22 11:19:51    941.0  2014-02-22 11:19:51  942.0   ...     \n",
       "2  14768.0  2013-12-16 16:40:19  14769.0  2013-12-16 16:40:19   37.0   ...     \n",
       "3    782.0  2014-03-28 10:53:12    782.0  2014-03-28 10:53:42  782.0   ...     \n",
       "4    175.0  2014-02-28 10:55:22    178.0  2014-02-28 10:55:23  177.0   ...     \n",
       "\n",
       "                 time6    site7                time7    site8  \\\n",
       "0                  NaN      NaN                  NaN      NaN   \n",
       "1  2014-02-22 11:19:51   3847.0  2014-02-22 11:19:52   3846.0   \n",
       "2  2013-12-16 16:40:19  14768.0  2013-12-16 16:40:20  14768.0   \n",
       "3  2014-03-28 10:54:42    782.0  2014-03-28 10:55:12    782.0   \n",
       "4  2014-02-28 10:55:59    175.0  2014-02-28 10:55:59    177.0   \n",
       "\n",
       "                 time8    site9                time9   site10  \\\n",
       "0                  NaN      NaN                  NaN      NaN   \n",
       "1  2014-02-22 11:19:52   1516.0  2014-02-22 11:20:15   1518.0   \n",
       "2  2013-12-16 16:40:21  14768.0  2013-12-16 16:40:22  14768.0   \n",
       "3  2014-03-28 10:55:42    782.0  2014-03-28 10:56:12    782.0   \n",
       "4  2014-02-28 10:55:59    177.0  2014-02-28 10:57:06    178.0   \n",
       "\n",
       "                time10  target  \n",
       "0                  NaN       0  \n",
       "1  2014-02-22 11:20:16       0  \n",
       "2  2013-12-16 16:40:24       0  \n",
       "3  2014-03-28 10:56:42       0  \n",
       "4  2014-02-28 10:57:11       0  \n",
       "\n",
       "[5 rows x 22 columns]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>session_id</th>\n",
       "      <th>site1</th>\n",
       "      <th>time1</th>\n",
       "      <th>site2</th>\n",
       "      <th>time2</th>\n",
       "      <th>site3</th>\n",
       "      <th>time3</th>\n",
       "      <th>site4</th>\n",
       "      <th>time4</th>\n",
       "      <th>site5</th>\n",
       "      <th>...</th>\n",
       "      <th>site6</th>\n",
       "      <th>time6</th>\n",
       "      <th>site7</th>\n",
       "      <th>time7</th>\n",
       "      <th>site8</th>\n",
       "      <th>time8</th>\n",
       "      <th>site9</th>\n",
       "      <th>time9</th>\n",
       "      <th>site10</th>\n",
       "      <th>time10</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>29</td>\n",
       "      <td>2014-10-04 11:19:53</td>\n",
       "      <td>35.0</td>\n",
       "      <td>2014-10-04 11:19:53</td>\n",
       "      <td>22.0</td>\n",
       "      <td>2014-10-04 11:19:54</td>\n",
       "      <td>321.0</td>\n",
       "      <td>2014-10-04 11:19:54</td>\n",
       "      <td>23.0</td>\n",
       "      <td>...</td>\n",
       "      <td>2211.0</td>\n",
       "      <td>2014-10-04 11:19:54</td>\n",
       "      <td>6730.0</td>\n",
       "      <td>2014-10-04 11:19:54</td>\n",
       "      <td>21.0</td>\n",
       "      <td>2014-10-04 11:19:54</td>\n",
       "      <td>44582.0</td>\n",
       "      <td>2014-10-04 11:20:00</td>\n",
       "      <td>15336.0</td>\n",
       "      <td>2014-10-04 11:20:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>782</td>\n",
       "      <td>2014-07-03 11:00:28</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-07-03 11:00:53</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-07-03 11:00:58</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-07-03 11:01:06</td>\n",
       "      <td>782.0</td>\n",
       "      <td>...</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-07-03 11:01:10</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-07-03 11:01:23</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-07-03 11:01:29</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-07-03 11:01:30</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-07-03 11:01:53</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>55</td>\n",
       "      <td>2014-12-05 15:55:12</td>\n",
       "      <td>55.0</td>\n",
       "      <td>2014-12-05 15:55:13</td>\n",
       "      <td>55.0</td>\n",
       "      <td>2014-12-05 15:55:14</td>\n",
       "      <td>55.0</td>\n",
       "      <td>2014-12-05 15:56:15</td>\n",
       "      <td>55.0</td>\n",
       "      <td>...</td>\n",
       "      <td>55.0</td>\n",
       "      <td>2014-12-05 15:56:17</td>\n",
       "      <td>55.0</td>\n",
       "      <td>2014-12-05 15:56:18</td>\n",
       "      <td>55.0</td>\n",
       "      <td>2014-12-05 15:56:19</td>\n",
       "      <td>1445.0</td>\n",
       "      <td>2014-12-05 15:56:33</td>\n",
       "      <td>1445.0</td>\n",
       "      <td>2014-12-05 15:56:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>1023</td>\n",
       "      <td>2014-11-04 10:03:19</td>\n",
       "      <td>1022.0</td>\n",
       "      <td>2014-11-04 10:03:19</td>\n",
       "      <td>50.0</td>\n",
       "      <td>2014-11-04 10:03:20</td>\n",
       "      <td>222.0</td>\n",
       "      <td>2014-11-04 10:03:21</td>\n",
       "      <td>202.0</td>\n",
       "      <td>...</td>\n",
       "      <td>3374.0</td>\n",
       "      <td>2014-11-04 10:03:22</td>\n",
       "      <td>50.0</td>\n",
       "      <td>2014-11-04 10:03:22</td>\n",
       "      <td>48.0</td>\n",
       "      <td>2014-11-04 10:03:22</td>\n",
       "      <td>48.0</td>\n",
       "      <td>2014-11-04 10:03:23</td>\n",
       "      <td>3374.0</td>\n",
       "      <td>2014-11-04 10:03:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>301</td>\n",
       "      <td>2014-05-16 15:05:31</td>\n",
       "      <td>301.0</td>\n",
       "      <td>2014-05-16 15:05:32</td>\n",
       "      <td>301.0</td>\n",
       "      <td>2014-05-16 15:05:33</td>\n",
       "      <td>66.0</td>\n",
       "      <td>2014-05-16 15:05:39</td>\n",
       "      <td>67.0</td>\n",
       "      <td>...</td>\n",
       "      <td>69.0</td>\n",
       "      <td>2014-05-16 15:05:40</td>\n",
       "      <td>70.0</td>\n",
       "      <td>2014-05-16 15:05:40</td>\n",
       "      <td>68.0</td>\n",
       "      <td>2014-05-16 15:05:40</td>\n",
       "      <td>71.0</td>\n",
       "      <td>2014-05-16 15:05:40</td>\n",
       "      <td>167.0</td>\n",
       "      <td>2014-05-16 15:05:44</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 21 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   session_id  site1                time1   site2                time2  site3  \\\n",
       "0           1     29  2014-10-04 11:19:53    35.0  2014-10-04 11:19:53   22.0   \n",
       "1           2    782  2014-07-03 11:00:28   782.0  2014-07-03 11:00:53  782.0   \n",
       "2           3     55  2014-12-05 15:55:12    55.0  2014-12-05 15:55:13   55.0   \n",
       "3           4   1023  2014-11-04 10:03:19  1022.0  2014-11-04 10:03:19   50.0   \n",
       "4           5    301  2014-05-16 15:05:31   301.0  2014-05-16 15:05:32  301.0   \n",
       "\n",
       "                 time3  site4                time4  site5  \\\n",
       "0  2014-10-04 11:19:54  321.0  2014-10-04 11:19:54   23.0   \n",
       "1  2014-07-03 11:00:58  782.0  2014-07-03 11:01:06  782.0   \n",
       "2  2014-12-05 15:55:14   55.0  2014-12-05 15:56:15   55.0   \n",
       "3  2014-11-04 10:03:20  222.0  2014-11-04 10:03:21  202.0   \n",
       "4  2014-05-16 15:05:33   66.0  2014-05-16 15:05:39   67.0   \n",
       "\n",
       "          ...            site6                time6   site7  \\\n",
       "0         ...           2211.0  2014-10-04 11:19:54  6730.0   \n",
       "1         ...            782.0  2014-07-03 11:01:10   782.0   \n",
       "2         ...             55.0  2014-12-05 15:56:17    55.0   \n",
       "3         ...           3374.0  2014-11-04 10:03:22    50.0   \n",
       "4         ...             69.0  2014-05-16 15:05:40    70.0   \n",
       "\n",
       "                 time7  site8                time8    site9  \\\n",
       "0  2014-10-04 11:19:54   21.0  2014-10-04 11:19:54  44582.0   \n",
       "1  2014-07-03 11:01:23  782.0  2014-07-03 11:01:29    782.0   \n",
       "2  2014-12-05 15:56:18   55.0  2014-12-05 15:56:19   1445.0   \n",
       "3  2014-11-04 10:03:22   48.0  2014-11-04 10:03:22     48.0   \n",
       "4  2014-05-16 15:05:40   68.0  2014-05-16 15:05:40     71.0   \n",
       "\n",
       "                 time9   site10               time10  \n",
       "0  2014-10-04 11:20:00  15336.0  2014-10-04 11:20:00  \n",
       "1  2014-07-03 11:01:30    782.0  2014-07-03 11:01:53  \n",
       "2  2014-12-05 15:56:33   1445.0  2014-12-05 15:56:36  \n",
       "3  2014-11-04 10:03:23   3374.0  2014-11-04 10:03:23  \n",
       "4  2014-05-16 15:05:40    167.0  2014-05-16 15:05:44  \n",
       "\n",
       "[5 rows x 21 columns]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def task_load_data():\n",
    "    train_df = pd.read_csv('data_5/train_sessions.csv')\n",
    "    display(train_df.head())\n",
    "    test_df = pd.read_csv('data_5/test_sessions.csv')\n",
    "    display(test_df.head())\n",
    "    \n",
    "    return Result(next_info=dict(train_df=train_df, test_df=test_df))\n",
    "\n",
    "task.call('load_data')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 5.1 Просмотр имеющихся признаков"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Сессии пользователей выделены таким образом, что они не могут быть длинее получаса или 10 сайтов. То есть сессия считается оконченной либо когда пользователь посетил 10 сайтов подряд, либо когда сессия заняла по времени более 30 минут. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Пропуски возникают там, где сессии короткие (менее 10 сайтов). Скажем, если человек 1 января 2015 года посетил *vk.com* в 20:01, потом *yandex.ru* в 20:29, затем *google.com* в 20:33, то первая его сессия будет состоять только из двух сайтов (site1 – ID сайта *vk.com*, time1 – 2015-01-01 20:01:00, site2 – ID сайта  *yandex.ru*, time2 – 2015-01-01 20:29:00, остальные признаки – NaN), а начиная с *google.com* пойдет новая сессия, потому что уже прошло более 30 минут с момента посещения *vk.com*.\n",
    "\n",
    "Посмотрим на признаки, возможные пропуски **(NaN).**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train_df.info():\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 253561 entries, 0 to 253560\n",
      "Data columns (total 22 columns):\n",
      "session_id    253561 non-null int64\n",
      "site1         253561 non-null int64\n",
      "time1         253561 non-null object\n",
      "site2         250098 non-null float64\n",
      "time2         250098 non-null object\n",
      "site3         246919 non-null float64\n",
      "time3         246919 non-null object\n",
      "site4         244321 non-null float64\n",
      "time4         244321 non-null object\n",
      "site5         241829 non-null float64\n",
      "time5         241829 non-null object\n",
      "site6         239495 non-null float64\n",
      "time6         239495 non-null object\n",
      "site7         237297 non-null float64\n",
      "time7         237297 non-null object\n",
      "site8         235224 non-null float64\n",
      "time8         235224 non-null object\n",
      "site9         233084 non-null float64\n",
      "time9         233084 non-null object\n",
      "site10        231052 non-null float64\n",
      "time10        231052 non-null object\n",
      "target        253561 non-null int64\n",
      "dtypes: float64(9), int64(3), object(10)\n",
      "memory usage: 42.6+ MB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "None"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train_df.head():\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>session_id</th>\n",
       "      <th>site1</th>\n",
       "      <th>time1</th>\n",
       "      <th>site2</th>\n",
       "      <th>time2</th>\n",
       "      <th>site3</th>\n",
       "      <th>time3</th>\n",
       "      <th>site4</th>\n",
       "      <th>time4</th>\n",
       "      <th>site5</th>\n",
       "      <th>...</th>\n",
       "      <th>time6</th>\n",
       "      <th>site7</th>\n",
       "      <th>time7</th>\n",
       "      <th>site8</th>\n",
       "      <th>time8</th>\n",
       "      <th>site9</th>\n",
       "      <th>time9</th>\n",
       "      <th>site10</th>\n",
       "      <th>time10</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>718</td>\n",
       "      <td>2014-02-20 10:02:45</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>890</td>\n",
       "      <td>2014-02-22 11:19:50</td>\n",
       "      <td>941.0</td>\n",
       "      <td>2014-02-22 11:19:50</td>\n",
       "      <td>3847.0</td>\n",
       "      <td>2014-02-22 11:19:51</td>\n",
       "      <td>941.0</td>\n",
       "      <td>2014-02-22 11:19:51</td>\n",
       "      <td>942.0</td>\n",
       "      <td>...</td>\n",
       "      <td>2014-02-22 11:19:51</td>\n",
       "      <td>3847.0</td>\n",
       "      <td>2014-02-22 11:19:52</td>\n",
       "      <td>3846.0</td>\n",
       "      <td>2014-02-22 11:19:52</td>\n",
       "      <td>1516.0</td>\n",
       "      <td>2014-02-22 11:20:15</td>\n",
       "      <td>1518.0</td>\n",
       "      <td>2014-02-22 11:20:16</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>14769</td>\n",
       "      <td>2013-12-16 16:40:17</td>\n",
       "      <td>39.0</td>\n",
       "      <td>2013-12-16 16:40:18</td>\n",
       "      <td>14768.0</td>\n",
       "      <td>2013-12-16 16:40:19</td>\n",
       "      <td>14769.0</td>\n",
       "      <td>2013-12-16 16:40:19</td>\n",
       "      <td>37.0</td>\n",
       "      <td>...</td>\n",
       "      <td>2013-12-16 16:40:19</td>\n",
       "      <td>14768.0</td>\n",
       "      <td>2013-12-16 16:40:20</td>\n",
       "      <td>14768.0</td>\n",
       "      <td>2013-12-16 16:40:21</td>\n",
       "      <td>14768.0</td>\n",
       "      <td>2013-12-16 16:40:22</td>\n",
       "      <td>14768.0</td>\n",
       "      <td>2013-12-16 16:40:24</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>782</td>\n",
       "      <td>2014-03-28 10:52:12</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-03-28 10:52:42</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-03-28 10:53:12</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-03-28 10:53:42</td>\n",
       "      <td>782.0</td>\n",
       "      <td>...</td>\n",
       "      <td>2014-03-28 10:54:42</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-03-28 10:55:12</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-03-28 10:55:42</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-03-28 10:56:12</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-03-28 10:56:42</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>22</td>\n",
       "      <td>2014-02-28 10:53:05</td>\n",
       "      <td>177.0</td>\n",
       "      <td>2014-02-28 10:55:22</td>\n",
       "      <td>175.0</td>\n",
       "      <td>2014-02-28 10:55:22</td>\n",
       "      <td>178.0</td>\n",
       "      <td>2014-02-28 10:55:23</td>\n",
       "      <td>177.0</td>\n",
       "      <td>...</td>\n",
       "      <td>2014-02-28 10:55:59</td>\n",
       "      <td>175.0</td>\n",
       "      <td>2014-02-28 10:55:59</td>\n",
       "      <td>177.0</td>\n",
       "      <td>2014-02-28 10:55:59</td>\n",
       "      <td>177.0</td>\n",
       "      <td>2014-02-28 10:57:06</td>\n",
       "      <td>178.0</td>\n",
       "      <td>2014-02-28 10:57:11</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 22 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   session_id  site1                time1  site2                time2  \\\n",
       "0           1    718  2014-02-20 10:02:45    NaN                  NaN   \n",
       "1           2    890  2014-02-22 11:19:50  941.0  2014-02-22 11:19:50   \n",
       "2           3  14769  2013-12-16 16:40:17   39.0  2013-12-16 16:40:18   \n",
       "3           4    782  2014-03-28 10:52:12  782.0  2014-03-28 10:52:42   \n",
       "4           5     22  2014-02-28 10:53:05  177.0  2014-02-28 10:55:22   \n",
       "\n",
       "     site3                time3    site4                time4  site5   ...    \\\n",
       "0      NaN                  NaN      NaN                  NaN    NaN   ...     \n",
       "1   3847.0  2014-02-22 11:19:51    941.0  2014-02-22 11:19:51  942.0   ...     \n",
       "2  14768.0  2013-12-16 16:40:19  14769.0  2013-12-16 16:40:19   37.0   ...     \n",
       "3    782.0  2014-03-28 10:53:12    782.0  2014-03-28 10:53:42  782.0   ...     \n",
       "4    175.0  2014-02-28 10:55:22    178.0  2014-02-28 10:55:23  177.0   ...     \n",
       "\n",
       "                 time6    site7                time7    site8  \\\n",
       "0                  NaN      NaN                  NaN      NaN   \n",
       "1  2014-02-22 11:19:51   3847.0  2014-02-22 11:19:52   3846.0   \n",
       "2  2013-12-16 16:40:19  14768.0  2013-12-16 16:40:20  14768.0   \n",
       "3  2014-03-28 10:54:42    782.0  2014-03-28 10:55:12    782.0   \n",
       "4  2014-02-28 10:55:59    175.0  2014-02-28 10:55:59    177.0   \n",
       "\n",
       "                 time8    site9                time9   site10  \\\n",
       "0                  NaN      NaN                  NaN      NaN   \n",
       "1  2014-02-22 11:19:52   1516.0  2014-02-22 11:20:15   1518.0   \n",
       "2  2013-12-16 16:40:21  14768.0  2013-12-16 16:40:22  14768.0   \n",
       "3  2014-03-28 10:55:42    782.0  2014-03-28 10:56:12    782.0   \n",
       "4  2014-02-28 10:55:59    177.0  2014-02-28 10:57:06    178.0   \n",
       "\n",
       "                time10  target  \n",
       "0                  NaN       0  \n",
       "1  2014-02-22 11:20:16       0  \n",
       "2  2013-12-16 16:40:24       0  \n",
       "3  2014-03-28 10:56:42       0  \n",
       "4  2014-02-28 10:57:11       0  \n",
       "\n",
       "[5 rows x 22 columns]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "test_df.info():\n",
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 82797 entries, 0 to 82796\n",
      "Data columns (total 21 columns):\n",
      "session_id    82797 non-null int64\n",
      "site1         82797 non-null int64\n",
      "time1         82797 non-null object\n",
      "site2         81308 non-null float64\n",
      "time2         81308 non-null object\n",
      "site3         80075 non-null float64\n",
      "time3         80075 non-null object\n",
      "site4         79182 non-null float64\n",
      "time4         79182 non-null object\n",
      "site5         78341 non-null float64\n",
      "time5         78341 non-null object\n",
      "site6         77566 non-null float64\n",
      "time6         77566 non-null object\n",
      "site7         76840 non-null float64\n",
      "time7         76840 non-null object\n",
      "site8         76151 non-null float64\n",
      "time8         76151 non-null object\n",
      "site9         75484 non-null float64\n",
      "time9         75484 non-null object\n",
      "site10        74806 non-null float64\n",
      "time10        74806 non-null object\n",
      "dtypes: float64(9), int64(2), object(10)\n",
      "memory usage: 13.3+ MB\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "None"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "test_df.head():\n",
      "\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>session_id</th>\n",
       "      <th>site1</th>\n",
       "      <th>time1</th>\n",
       "      <th>site2</th>\n",
       "      <th>time2</th>\n",
       "      <th>site3</th>\n",
       "      <th>time3</th>\n",
       "      <th>site4</th>\n",
       "      <th>time4</th>\n",
       "      <th>site5</th>\n",
       "      <th>...</th>\n",
       "      <th>site6</th>\n",
       "      <th>time6</th>\n",
       "      <th>site7</th>\n",
       "      <th>time7</th>\n",
       "      <th>site8</th>\n",
       "      <th>time8</th>\n",
       "      <th>site9</th>\n",
       "      <th>time9</th>\n",
       "      <th>site10</th>\n",
       "      <th>time10</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>29</td>\n",
       "      <td>2014-10-04 11:19:53</td>\n",
       "      <td>35.0</td>\n",
       "      <td>2014-10-04 11:19:53</td>\n",
       "      <td>22.0</td>\n",
       "      <td>2014-10-04 11:19:54</td>\n",
       "      <td>321.0</td>\n",
       "      <td>2014-10-04 11:19:54</td>\n",
       "      <td>23.0</td>\n",
       "      <td>...</td>\n",
       "      <td>2211.0</td>\n",
       "      <td>2014-10-04 11:19:54</td>\n",
       "      <td>6730.0</td>\n",
       "      <td>2014-10-04 11:19:54</td>\n",
       "      <td>21.0</td>\n",
       "      <td>2014-10-04 11:19:54</td>\n",
       "      <td>44582.0</td>\n",
       "      <td>2014-10-04 11:20:00</td>\n",
       "      <td>15336.0</td>\n",
       "      <td>2014-10-04 11:20:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>782</td>\n",
       "      <td>2014-07-03 11:00:28</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-07-03 11:00:53</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-07-03 11:00:58</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-07-03 11:01:06</td>\n",
       "      <td>782.0</td>\n",
       "      <td>...</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-07-03 11:01:10</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-07-03 11:01:23</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-07-03 11:01:29</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-07-03 11:01:30</td>\n",
       "      <td>782.0</td>\n",
       "      <td>2014-07-03 11:01:53</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>55</td>\n",
       "      <td>2014-12-05 15:55:12</td>\n",
       "      <td>55.0</td>\n",
       "      <td>2014-12-05 15:55:13</td>\n",
       "      <td>55.0</td>\n",
       "      <td>2014-12-05 15:55:14</td>\n",
       "      <td>55.0</td>\n",
       "      <td>2014-12-05 15:56:15</td>\n",
       "      <td>55.0</td>\n",
       "      <td>...</td>\n",
       "      <td>55.0</td>\n",
       "      <td>2014-12-05 15:56:17</td>\n",
       "      <td>55.0</td>\n",
       "      <td>2014-12-05 15:56:18</td>\n",
       "      <td>55.0</td>\n",
       "      <td>2014-12-05 15:56:19</td>\n",
       "      <td>1445.0</td>\n",
       "      <td>2014-12-05 15:56:33</td>\n",
       "      <td>1445.0</td>\n",
       "      <td>2014-12-05 15:56:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>1023</td>\n",
       "      <td>2014-11-04 10:03:19</td>\n",
       "      <td>1022.0</td>\n",
       "      <td>2014-11-04 10:03:19</td>\n",
       "      <td>50.0</td>\n",
       "      <td>2014-11-04 10:03:20</td>\n",
       "      <td>222.0</td>\n",
       "      <td>2014-11-04 10:03:21</td>\n",
       "      <td>202.0</td>\n",
       "      <td>...</td>\n",
       "      <td>3374.0</td>\n",
       "      <td>2014-11-04 10:03:22</td>\n",
       "      <td>50.0</td>\n",
       "      <td>2014-11-04 10:03:22</td>\n",
       "      <td>48.0</td>\n",
       "      <td>2014-11-04 10:03:22</td>\n",
       "      <td>48.0</td>\n",
       "      <td>2014-11-04 10:03:23</td>\n",
       "      <td>3374.0</td>\n",
       "      <td>2014-11-04 10:03:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>301</td>\n",
       "      <td>2014-05-16 15:05:31</td>\n",
       "      <td>301.0</td>\n",
       "      <td>2014-05-16 15:05:32</td>\n",
       "      <td>301.0</td>\n",
       "      <td>2014-05-16 15:05:33</td>\n",
       "      <td>66.0</td>\n",
       "      <td>2014-05-16 15:05:39</td>\n",
       "      <td>67.0</td>\n",
       "      <td>...</td>\n",
       "      <td>69.0</td>\n",
       "      <td>2014-05-16 15:05:40</td>\n",
       "      <td>70.0</td>\n",
       "      <td>2014-05-16 15:05:40</td>\n",
       "      <td>68.0</td>\n",
       "      <td>2014-05-16 15:05:40</td>\n",
       "      <td>71.0</td>\n",
       "      <td>2014-05-16 15:05:40</td>\n",
       "      <td>167.0</td>\n",
       "      <td>2014-05-16 15:05:44</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 21 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   session_id  site1                time1   site2                time2  site3  \\\n",
       "0           1     29  2014-10-04 11:19:53    35.0  2014-10-04 11:19:53   22.0   \n",
       "1           2    782  2014-07-03 11:00:28   782.0  2014-07-03 11:00:53  782.0   \n",
       "2           3     55  2014-12-05 15:55:12    55.0  2014-12-05 15:55:13   55.0   \n",
       "3           4   1023  2014-11-04 10:03:19  1022.0  2014-11-04 10:03:19   50.0   \n",
       "4           5    301  2014-05-16 15:05:31   301.0  2014-05-16 15:05:32  301.0   \n",
       "\n",
       "                 time3  site4                time4  site5  \\\n",
       "0  2014-10-04 11:19:54  321.0  2014-10-04 11:19:54   23.0   \n",
       "1  2014-07-03 11:00:58  782.0  2014-07-03 11:01:06  782.0   \n",
       "2  2014-12-05 15:55:14   55.0  2014-12-05 15:56:15   55.0   \n",
       "3  2014-11-04 10:03:20  222.0  2014-11-04 10:03:21  202.0   \n",
       "4  2014-05-16 15:05:33   66.0  2014-05-16 15:05:39   67.0   \n",
       "\n",
       "          ...            site6                time6   site7  \\\n",
       "0         ...           2211.0  2014-10-04 11:19:54  6730.0   \n",
       "1         ...            782.0  2014-07-03 11:01:10   782.0   \n",
       "2         ...             55.0  2014-12-05 15:56:17    55.0   \n",
       "3         ...           3374.0  2014-11-04 10:03:22    50.0   \n",
       "4         ...             69.0  2014-05-16 15:05:40    70.0   \n",
       "\n",
       "                 time7  site8                time8    site9  \\\n",
       "0  2014-10-04 11:19:54   21.0  2014-10-04 11:19:54  44582.0   \n",
       "1  2014-07-03 11:01:23  782.0  2014-07-03 11:01:29    782.0   \n",
       "2  2014-12-05 15:56:18   55.0  2014-12-05 15:56:19   1445.0   \n",
       "3  2014-11-04 10:03:22   48.0  2014-11-04 10:03:22     48.0   \n",
       "4  2014-05-16 15:05:40   68.0  2014-05-16 15:05:40     71.0   \n",
       "\n",
       "                 time9   site10               time10  \n",
       "0  2014-10-04 11:20:00  15336.0  2014-10-04 11:20:00  \n",
       "1  2014-07-03 11:01:30    782.0  2014-07-03 11:01:53  \n",
       "2  2014-12-05 15:56:33   1445.0  2014-12-05 15:56:36  \n",
       "3  2014-11-04 10:03:23   3374.0  2014-11-04 10:03:23  \n",
       "4  2014-05-16 15:05:40    167.0  2014-05-16 15:05:44  \n",
       "\n",
       "[5 rows x 21 columns]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train_df['target'].value_counts():\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0    251264\n",
       "1      2297\n",
       "Name: target, dtype: int64"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "@use_result('load_data') # это и задает pipe - от какой задачи принимаем данные (см. next_info пред. задачи)\n",
    "def task_info(train_df, test_df, **kwargs):\n",
    "    print('train_df.info():')\n",
    "    display(train_df.info())\n",
    "    print('train_df.head():')\n",
    "    display(train_df.head())\n",
    "    print('test_df.info():')\n",
    "    display(test_df.info())\n",
    "    print('test_df.head():\\n')\n",
    "    display(test_df.head())\n",
    "    print('train_df[\\'target\\'].value_counts():')\n",
    "    display(train_df['target'].value_counts())\n",
    "\n",
    "    return Result(next_info=dict(nrows_train_df=train_df.shape[0], nrows_test_df=test_df.shape[0]))\n",
    "\n",
    "task.call('info')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 5.2. Получение новых признаков"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "В ходе задачи было установлено, что с помощью имеющихся признаков сложно получить модель приемлимого качества. Поэтому из исходных данных можно получить дополнительные признаки. \n",
    "\n",
    "Синтезируем новые признаки для анализа. Данные синтезируются с помощью 2 классов: *SitesData2* (вторая версия класса, изначально использованного в 3 неделе курса) и *FeatureMaker.*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def task_my_create_data( \\\n",
    "        path_to_train_csv, \n",
    "        path_to_test_csv, \n",
    "        path_to_site_dic, \n",
    "        path_train_features_pickle, \n",
    "        path_train_targets_pickle,\n",
    "        path_test_features_pickle,\n",
    "        path_train_site_cnts_pickle,\n",
    "        path_test_site_cnts_pickle):\n",
    "\n",
    "    sites_data2 = SitesData2( \\\n",
    "        path_to_train_csv=path_to_train_csv,\n",
    "        path_to_test_csv=path_to_test_csv,\n",
    "        path_to_site_dic=path_to_site_dic)\n",
    "    \n",
    "    sites_data2.prepare_train_features()\n",
    "    sites_data2.pickle_train_features(path_train_features_pickle)    \n",
    "    sites_data2.pickle_train_targets(path_train_targets_pickle)\n",
    "    sites_data2.prepare_test_features()\n",
    "    sites_data2.pickle_test_features(path_test_features_pickle)    \n",
    "    sites_data2.prepare_site_cnts()\n",
    "    sites_data2.pickle_train_site_cnts(path_train_site_cnts_pickle)\n",
    "    sites_data2.pickle_test_site_cnts(path_test_site_cnts_pickle)\n",
    "    \n",
    "task.call( \\\n",
    "    'my_create_data', \n",
    "    use_result=Result(next_info=dict( \\\n",
    "            path_to_train_csv='data_5/train_sessions.csv',\n",
    "            path_to_test_csv='data_5/test_sessions.csv', \n",
    "            path_to_site_dic='data/site_dic.pkl', \n",
    "            path_train_features_pickle='result/my_train_features.pkl', \n",
    "            path_train_targets_pickle='result/my_train_targets.pkl', \n",
    "            path_test_features_pickle='result/my_test_features.pkl',\n",
    "            path_train_site_cnts_pickle='result/my_train_site_cnts.pkl',\n",
    "            path_test_site_cnts_pickle='result/my_test_site_cnts.pkl'\n",
    "        )\n",
    "    )\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Задача загрузки сгенерированных признаков, чтобы не генерировать их заново (это может быть долго), если случился перерыв в работе. В случае перерыва выполнять предыдущие задачи не обязательно, эта задача не связана по *pipe* с результатом предыдущих (не используется декоратор **@use_result()).** Достаточно выполнить ячейки инициализации библиотек и создания переменной ***task,*** что я и сделал."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "def task_my_load_data( \\\n",
    "    path_train_features_pickle, \n",
    "    path_train_targets_pickle, \n",
    "    path_test_features_pickle,\n",
    "    path_train_site_cnts_pickle,\n",
    "    path_test_site_cnts_pickle):\n",
    "\n",
    "    train_features = SitesData2.unpickle(path_train_features_pickle)\n",
    "    train_targets = SitesData2.unpickle(path_train_targets_pickle)\n",
    "    test_features = SitesData2.unpickle(path_test_features_pickle)\n",
    "    train_site_cnts = SitesData2.unpickle(path_train_site_cnts_pickle)\n",
    "    test_site_cnts = SitesData2.unpickle(path_test_site_cnts_pickle)\n",
    "    \n",
    "    elice_indices = np.where(train_targets==1)[0]\n",
    "    other_indices = np.where(train_targets!=1)[0]\n",
    "\n",
    "    return Result(next_info=dict( \\\n",
    "        train_features=train_features,\n",
    "        train_targets=train_targets,\n",
    "        test_features=test_features,\n",
    "        train_site_cnts=train_site_cnts,\n",
    "        test_site_cnts=test_site_cnts,\n",
    "        elice_indices=elice_indices,\n",
    "        other_indices=other_indices))\n",
    "\n",
    "task.call('my_load_data', use_result=Result(next_info=dict( \\\n",
    "                path_train_features_pickle='result/my_train_features.pkl', \n",
    "                path_train_targets_pickle='result/my_train_targets.pkl', \n",
    "                path_test_features_pickle='result/my_test_features.pkl',\n",
    "                path_train_site_cnts_pickle='result/my_train_site_cnts.pkl',\n",
    "                path_test_site_cnts_pickle='result/my_test_site_cnts.pkl'\n",
    "            )\n",
    "        )\n",
    "    )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 5.3. Формальный анализ признаков"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Медианная, минимальная, максимальная продолжительсть сессии для целевого пользователя Элис и всех остальных пользователей."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Убедимся, что классов всего 2, выведем уникальные классы: [0 1]\n",
      "Медианная продолжительность сессии Элис: 11.0\n",
      "Медианная продолжительность сессии других пользователей: 28.0\n",
      "Минимальная продолжительность сессии: 0\n",
      "Максимальная продолжительность сессии: 1800\n"
     ]
    }
   ],
   "source": [
    "@use_result('my_load_data')\n",
    "def task_median_session_timespan(train_features, train_targets, elice_indices, other_indices, **kwargs):\n",
    "    print('Убедимся, что классов всего 2, выведем уникальные классы:', np.unique(train_targets))\n",
    "    print('Медианная продолжительность сессии Элис:', train_features.iloc[elice_indices]['session_timespan'].median())\n",
    "    print('Медианная продолжительность сессии других пользователей:', train_features.iloc[other_indices]['session_timespan'].median())\n",
    "    print('Минимальная продолжительность сессии:', train_features['session_timespan'].min())\n",
    "    print('Максимальная продолжительность сессии:', train_features['session_timespan'].max())\n",
    "    \n",
    "task.call('median_session_timespan')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Всё правильно, в описании данных было сказано, что сессии длиннее чем полчаса (1800 секунд) обрезаются. В то же время, если в сессии только два сайта и интервал между их посещением менее чем 1 секунда — то сессия имеет нулевую длину в секундах."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Медианный день недели. Также выведем минимальный и максимальный день, чтобы увидеть, с какого числа нумеруются дни - с ноля или с единицы."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Медианный день недели Элис: 1.0\n",
      "Медианный день недели других пользователей: 2.0\n",
      "Номер первого дня в неделе: 0\n",
      "Номер последнего дня в неделе: 6\n"
     ]
    }
   ],
   "source": [
    "@use_result('my_load_data')\n",
    "def task_median_session_dayofweek(train_features, train_targets, elice_indices, other_indices, **kwargs):\n",
    "    # 1 - вторник, второй день\n",
    "    print('Медианный день недели Элис:', train_features.iloc[elice_indices]['day_of_week'].median())\n",
    "    # 2 - среда, третий день\n",
    "    print('Медианный день недели других пользователей:', train_features.iloc[other_indices]['day_of_week'].median())\n",
    "    print('Номер первого дня в неделе:', train_features['day_of_week'].min())\n",
    "    print('Номер последнего дня в неделе:', train_features['day_of_week'].max())\n",
    "    \n",
    "task.call('median_session_dayofweek')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "Медианный час начала сессии."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Медианный час начала сессии Элис: 16.0\n",
      "Медианный час начала сессии других пользователей: 12.0\n"
     ]
    }
   ],
   "source": [
    "@use_result('my_load_data')\n",
    "def task_median_session_starthour(train_features, train_targets, elice_indices, other_indices, **kwargs):\n",
    "    print('Медианный час начала сессии Элис:', train_features.iloc[elice_indices]['start_hour'].median())\n",
    "    print('Медианный час начала сессии других пользователей:', train_features.iloc[other_indices]['start_hour'].median())\n",
    "    \n",
    "task.call('median_session_starthour')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Медианное значение числа уникальных сайтов."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Медианное число уникальных сайтов в сессии Элис: 6.0\n",
      "Медианное число уникальных сайтов в сессии других пользователей: 6.0\n"
     ]
    }
   ],
   "source": [
    "@use_result('my_load_data')\n",
    "def task_median_session_uniquesites(train_features, train_targets, elice_indices, other_indices, **kwargs):\n",
    "    print('Медианное число уникальных сайтов в сессии Элис:', train_features.iloc[elice_indices]['#unique_sites'].median())\n",
    "    print('Медианное число уникальных сайтов в сессии других пользователей:', train_features.iloc[other_indices]['#unique_sites'].median())\n",
    "    \n",
    "task.call('median_session_uniquesites')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 6. Визуальный анализ исходных данных и признаков"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "Посмотрим на разность распределения признаков Элис и других пользователей."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 6.1. Сравнение распределений некоторых признаков."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "@use_result('my_load_data')\n",
    "def task_hist_data(train_features, elice_indices, other_indices, **kwargs):\n",
    "    \n",
    "    color_dict = {1: 'magenta', 0: 'gray'} # для Элис розовый цвет, для других — серый\n",
    "    usernames = {1: u'Элис', 0: u'Другие пользователи'}\n",
    "    \n",
    "    features_by_user = {1: train_features.iloc[elice_indices], 0: train_features.iloc[other_indices]}\n",
    "\n",
    "    return Result(next_info=dict(color_dict=color_dict, usernames=usernames, features_by_user=features_by_user))\n",
    "\n",
    "task.call('hist_data')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Распределение продолжительности сессии Элис и других пользователей."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABHgAAALICAYAAAAE6EcMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1fb953d190>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "@use_result('hist_data')\n",
    "def task_hist_session_timespan(features_by_user, color_dict, usernames, **kwargs):\n",
    "    hist_params = []\n",
    "\n",
    "    for target, username in usernames.iteritems():\n",
    "        plot_data = features_by_user[target]['session_timespan']\n",
    "\n",
    "        hist_params.append(\n",
    "            HistParams(\n",
    "                data=plot_data, \n",
    "                title=u'Распределение длины сессии',\n",
    "                color=color_dict[target],\n",
    "                xticklabels=np.linspace(0, 1800, 31),\n",
    "                legend=username\n",
    "            )\n",
    "        )\n",
    "\n",
    "    color_hist(u'Длина сессии', u'Количество', *hist_params, rotation=90)\n",
    "    \n",
    "task.call('hist_session_timespan')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "Соответсвующие графики для числа уникальных сайтов в сессии — для Элис и других пользователей."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABHgAAALJCAYAAADPtJSpAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nWNq0CAut0+0AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1fb9686210>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "@use_result('hist_data')\n",
    "def task_hist_unique_sites(features_by_user, color_dict, usernames, **kwargs):\n",
    "    hist_params = []\n",
    "\n",
    "    for target, username in usernames.iteritems():\n",
    "        plot_data = features_by_user[target]['#unique_sites']\n",
    "\n",
    "        hist_params.append(\n",
    "            HistParams(\n",
    "                data=plot_data, \n",
    "                title=u'Распределение числа уникальных сайтов',\n",
    "                color=color_dict[target],\n",
    "                xticklabels=range(1, 11),\n",
    "                legend=username\n",
    "            )\n",
    "        )\n",
    "\n",
    "    color_hist(u'Число уникальных сайтов', u'Количество', *hist_params)\n",
    "    \n",
    "task.call('hist_unique_sites')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Распределение часа начала сессии."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABHgAAALJCAYAAADPtJSpAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1fb986e650>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "@use_result('hist_data')\n",
    "def task_hist_start_hour(features_by_user, color_dict, usernames, **kwargs):\n",
    "    hist_params = []\n",
    "\n",
    "    for target, username in usernames.iteritems():\n",
    "        plot_data = features_by_user[target]['start_hour']\n",
    "\n",
    "        hist_params.append(\n",
    "            HistParams(\n",
    "                data=plot_data, \n",
    "                title=u'Распределение часа начала сессии',\n",
    "                color=color_dict[target],\n",
    "                xticklabels=range(0, 24),\n",
    "                legend=username\n",
    "            )\n",
    "        )\n",
    "\n",
    "    color_hist(u'Час начала сессии', u'Количество', *hist_params)\n",
    "    \n",
    "task.call('hist_start_hour')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "День недели, в который началась сессия."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABHgAAALJCAYAAADPtJSpAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzs3Xl8XXWd//FXWkqhDSVCSgsV6SJ+2EZHwR9SwEoRHOfn\nNhRkBJkBZLUsVnQGBWRRGUbcAEEQRRCVAZwqMvpDdASFEWEUlUX8CJRUpaUkdqFFW9o0vz/OKaQh\nbW9obm5O83r64MG9557l8z1pycf3Ped7mrq6upAkSZIkSVJ1DWt0AZIkSZIkSdo4BjySJEmSJEkV\nZ8AjSZIkSZJUcQY8kiRJkiRJFWfAI0mSJEmSVHEGPJIkSZIkSRW3WaMLkDZFEbEaeAxYBQwHFgMf\nycwfN7QwSZKkPrCnkaTq8AoeqT66gGmZuVtmBjALuDkitm1wXZIkSX1hTyNJFeEVPFJ9NJX/AJCZ\nP4uIx4B9gP+KiOOAD1J8EzYfOCoz/wgQEZ8F3gU8B3w5Mz8dEf8MXA78odzv9sCnMvPCiHgCuAI4\nHNgRuCozP1bu653Ax4FRFN++HZGZC8vPpgI/AP4IjAW+nZknlJ99DDgCGAl8B/hgZnaVn30P+Bvg\nWaB1zXYRsTVwGbB3Oa5PZOa15TargZdn5ryI2Al4LDNHRMQ04LzMPKBc71vANpk5PSI2Bz4NvAUY\nAVydmf/W80RHxFeBRzPzwl6O1dfz3AR8ATiwPOb/AMdkZmePY04rt9m5fP/PwHsz86CI2BK4FnhN\nuY/Zmfnhcr0ngCPLPw/P72NDx42ILwAzgEVAC/BQZh68vmP1cp5eNN71/awjYlK57x2AhcBJmfmr\n9Sy/Azg3M38aEX8D/Bo4oHy/OjOHlcc7Bbh0zXtJ0qBnT2NPY09jT6OK8A+jNHBGACsiYixwKXBg\n+U3Y48A5ABHxXmAv4JXA64FTImKvcvufld+e7Qp8u8e+35CZewF7ADMj4m/KX1pfAw7PzFcCdwBX\nddtmIvCdzNyN4hcxZQ1HAYeWdUwp/zm5x3aHlHV8odvyzwKd5Zj2Bs6PiN3Kz7p61NvV83VEvBqY\n1m35vwK7ALuX/8yIiL9nw9bs76Wc538A9gV2A3YF9qRoMtd5nF7enwyMzsxdgNcBR5eN5/r2saHj\nTgJOLX9WH+m2vKZjRcSRvY13Az/rLwHfKBu+Cyn+LK1veXfnUTRKa40zIkYCJ/DicydJqhZ7mhfY\n06y9jT2N1EAGPNIAiIi3AuOA/8nMdmBMZs4vP74LmFy+fivwrcxcnZlLgV0z8xc1HOJrAOW+7wKm\nAn8H3JGZj5TrfAl4R/nNCkAAj/ayr7cB12TmssxcDXyF4puWNSYCbevY7pKyjj8Ds4FDys+auq3X\nRO/OBS7usb8rMnNVZv4VuL7b/rrr+Yu1qayhz+c5M2cDe5XLnwP+t9s269P9m8013yqRmUuAh7vt\no9cmoIbjTqKXc76BY3X3972Nl95/1oeUTcsBwH+U+74F2Htdy7sfqGxqtwQe7KWOE4Cv93YOJEnV\nYE9jT7OOWtdsY08jNZC3aEn1c2dErKIIUtuAt2bmXyJiGPCJiHh7+dkYIMttWikmLwSgbAJq0f2b\nhUXAyyh+QU+LiN+Wy5vKz7YFOigugb2ul321AB+KiBPKbYYDTwNExA7AiszsWMd2N5VjbgK2AG7q\n9vma87F5zw0j4jXlce6laOLW7O9zEXFhub/Ny8976qD4FoeIeDkvfLPS5/McEa3AZRHxOmA1RQP7\n+V6OCbBTeW6bgK2Bh8p9vBL4bEREuY+XA9f0qPVnFJeerxn/Oo9bNq+T6KVx3cCxulvXn6t1/ay3\nAZoy85lu2/wlIrbvbXm34zRRNLVnA5/pUetI4GhgP+CiXmqUJA1e9jT2NPY0L9RqT6NBy4BHqp9p\n3b5p6e5wim8Z9svMReU91UeUn3VQ/OICICK2A2ppiFq7vd6GojlaAfwwM9/dc+XyF+xU4NRe9jUP\nuCUzr+jls12A3/ayfM1278rMdX0+LTPnR3G/evdf7E0Ulxl/jKKJ676/izPz++vY3xpfpJjs8QHW\nbpb+kb6f509S3M+9e2auioj1fTMzt7y8eM396keWyy8HfpGZ7yg/u7vbNucDn4+I9wN/6rb8wvUc\ndxLw5/LbrJ7Wd6zu1jXeXn/WUcwV0BUR2+QL8xtMoZjb4EXLM/PxctO/Bboy8/6iP3teE3AicG1m\n/jUivJxZkqrFnmZt9jT2NPY0GpS8RUuqn3Vdtrsd0Fb+gt4WeDfQXH72XeA9EbF5RIwG7qa4V3tD\nDo+IpogYR3Hf808pJhvcP4r71omI/xMRa7652Q9YnOXkfD3cAhwVxWR3RMQJ5X3NAAcBP1lHDbdQ\n3uscEZtFxGcj4m+7fb6uS5p3AZ7LzAd62d/xETGsHNtZEXFwz4Nm5hOZuVdmvjozj++277H0/Txv\nBzxYNiSvoTiXzfTNdsCvyvNwEMW3W81lrd/LzJ0z8w0UkzeuMXY9xz2YdZ/zdR6rh3WNt9efdXlJ\n9e0U304REX8HfG9dy7sd5wzKOQF68R7gyvL1uv5uSJIGJ3saexp7mhfY02jQMuCR6mN9af4NQGtE\n/B74BnAWsGNEXJyZN1L8snkU+CXFUxZ+XsPxHgbuo7hH+JLM/F1mPgUcD3w7Ih6mmJzvPyLi9cB/\nAttFxG8j4hFgJnBoRLw/M78D/Bdwf3m57tuBH0TEhyjuje7tWzAofgluHRG/K+sYBqxpcNY3IWEr\nxQR2PV0OzC3H9luKpmld3+b0tu+Xcp4/A5xcnq+TKZ5WcVxEzOh5kPX4BMUlxg8A+1N8w3V+ROyz\nnm3Wddz3UkzM+KmNOVY53h/0HG/5s76VHj/rcrPjKeY3eBy4gKKZWd9ygJ90mx+h56STX87Mlb18\nJkka3Oxp7GnsaV5gT6NBramrq75/JqOY6fzDwEqKyxUfpJhYbBgvPOJvZbne6UAnxV/UayJiM4pH\n1+0ErKJ4xF5bFBNefZHi/swHMnNmXQchDWLR7TGVNa4/jeLRj9N7LD+X4lLUC+pQpiQNGVHcMnIl\nxVOAVgAnAX+hxv6nMVVLjWdPI0kbp65X8ETENhShzlSK+0bfRZGOXpaZ0yge8XdsRIyiSMqnU8xm\nPisiWijuLV2UmftT3M+5ZhKrz1M8Xm9/oCUi3lLPcUiSJPXBOymeeLMv8D6Kb7T70v9IkiT1Wb1v\n0XozxYRof8nMBZl5IvAmisvnKP99EMUj6e7L4rF2yykuWdwPOBD4drnuj4CpETECmJSZ93fbx5vr\nPA5pMPPSUEkaXHamuMWEzHyC4krkadTW/+w78OVKg4Y9jSRthHo/RWsiMDoibqF4dN35wKhu9yw+\nDWxP8fi89m7btfdcnpldUcxSPp61H5+4Zh/SkJSZk/u4/k8ovi3uufz8fitKkoa2B4EPRMQlFGHP\nZGDLPvQ/0pBkTyNJG6feAU8TxeMN/4Ei7LmDdc8633O7dS3vqnEfz3vkkezq7Fy5odUkSdImao89\n9hiwJ51k5m0RMZXiaTEPAI8Af9Ntlb72P8+zp5EkaWhbX09T74BnAfCzzFwNzImIpcDKiBiZmSuA\nCcCTwDzW/sZqAnBPuXw88GA54XITxcSE2/ZYd976imht3aGfhlNdY8duRXv70kaX0TCOf2iPHzwH\njn9ojx88BwMtMz+25nVEPAb8qQ/9zzrZ0/hn2fEP7fGD58DxO/6hPP4NqfccPLcD0yOiKSK2BZop\n5tI5tPx8BnAbxX3qe0XEmIhoppiU+S7gh8Bh5brvAO7IzE7gkfKbMYBDyn1IkiQ1XES8OiK+Ur7+\nO4pH+fal/5EkSeqzul7Bk5nzIuJbwM8pbq2aCfwCuD4iTgDmAtdlZmdEnEkRCK0GzsvMpRFxI3BQ\nRNwFLAeOLnc9C7iqfAzpvZn543qOQ5IkqQ8eBJoi4l7gr8CRFI9B/1ot/U+jipYkSdVW71u0yMyr\ngat7LD64l/VmA7N7LFsNHNvLuo8Ab+zHMiVJkvpFZnbRS/9Cjf2PJEnSS1HvW7QkSZIkSZJUZwY8\nkiRJkiRJFWfAI0mSJEmSVHEGPJIkSZIkSRVX90mWJUmqp87OTtra5qx3nUWLmlm4cFnN+5w4cTLD\nhw/f2NIkSZJUJ7X0gH1V9R7QgEeSVGltbXO4/PLP0dLS0i/7W7x4MTNnzmLKlJ03uO7JJx/L8OGb\nsXLlSlav7uTqq7/WLzVIkiRp/dra5rBkn3YmMalf9vcET9B2DxvsAf/rv27hu9/9NiNGjGDVqlXM\nmPFuDj74rf1Sw8Yy4JEkVV5LSwutra0DftxFixbzH/8xm6eems855/zrgB9fkiRpKJvEJF7Fq/pt\nfwtZ/xXff/jDXG6//f9x6aVXssUWW/DXv/6Vj3zkDF7+8h3Zbbc9+q2Ol8qAR5Kkl2D16tU0Nb14\n+TXXfIn29qdZsOAp/vznP/P+95/G/PlP0tHRwXHHnQTArFkzOeWUWZx55gcZN248Eya8nGOOOZ5P\nfvI8LrvsKv7wh7kcddS7ufHGWxgzZgwXXng+y5YtpbOzk1mzPszkya9c65jTpu3Na17zWrq6unj4\n4Yf4/vf/m6VLn+Hf/u0CVq5cybBhw7n44ovYfPMxPPro7znjjFOZOHES8+fP4z3vOYqpU/fj7LP/\nlS9/ee0rkM4//2yefnoBy5cv5yMf+RgTJ07iU5/6JPPmPcmqVat43/tO5PWv35tTTz2R5cuXM3Lk\nSEaMGMFFF32GZ555ho9//GM0NTWxatUqzjrrPB577FFuvvkGHn309+y444684hU7cdZZ5/OpT32S\n+fPnPb/P171uLwCOOeYImpu34rnnnnv+CqlTTz2Rs846n/Hjx3PKKSew556v55hjjq/PD1mSJKmb\nV7xiJy699Mrn32+55ZYcc8wJ3HjjN3nve4/utcfqrb9rbm7mggvO5tlnn6W5uZnzz/83tthii42u\nz4BHkqSXoL39aVpbx/b6WUdHO5/97BeYM+cxPvGJc7nkkis55ZTjOe64k3j22WU888wzTJlShDSf\n+cyljBy5BU89NZ+mMjH6xjeuY7vtxgNw003f5A1vmMrb3vZO2tqe4JJLPs3nPnf5Wsfbaqsxzzcb\nhx32TgC+/OUredvb3sX06W/mzjv/m8suu4wzzjiLBQvmc+CBB3P66WdwzTVfen4fPcOqVatW8a53\nHcprXvO3XHvtl7n33p/x2GO/Z+TILfjCF75ER0cHp556AjfcMBuAj370XCZNmsxpp53E/PnzWb78\nrxxzzPG89rV78r3vfZfZs2/mlFM+wBvf+CZOO+0kPvjBf2XixEn84Affp7V1LGeeeQ5LlizmtNNO\n5rrrbgBg2bJlfPWr3+z1Cqn77/8FHR0dff65SZIkbaxHHnmYK664lKamJlaufI4VK55bT4/14v7u\nhhuuZ++992HGjMO56aYb+MUv7mW//aZtdF0GPJIkvQRPPvkndthhQq+f7bnn6wGYPPmVdHR0sNVW\nW7Hjjq/g0UeTuXPbOOCAA59ft6tr7W3nzXuSYcOGs/322wPw0EMPsGTJYn7wg+8D8Nxzz/VyxLV3\n0tXVxe9+9wgnnXQqAK973V58/etfLfc/r9e6586dy2mnncTq1as5/PAj2H//N/Ga1/wtM2cezzPP\nLOHKK6/h6quv5LWv3ROA1tZWNt98JEuXLgXgwgvPY+XKVeywwwQmTpzE008v4POf/zRf+cpVLF26\nlIhd1qqvqxz4gw8+wIMP/poHHvg1XV1drFz5HKtWrWL58uVstdVWvZ5fgP/8zxuZMePdLFu2dJ3r\nSJIk1cOuu+7OZZddBcDvf/87/v3fP8n8+fPX2Ruu3d918fvf/47jjz8ZgHe/+z39VpcBjyRJL8Hv\nfvfb56/C6Wn16q4XLXvLW/6eH//4RyxY8BQnnDCz1+26urq44Yavc9RRR3PRRR8HYMSIEXzgA//C\n7ruv777utS+/aWpqKr8tKuoobtMaBsC8eX/iDW/Y90V72Gmn4pLjRYsW8b73vZf9938TAJdffjVX\nXHEps2ffzLBhTc8HMwCrVq1k2LDi2GeddT4TJ07ioos+zl133cldd/2Evffeh3e+8xDuvPO/+dnP\n7u618hEjRvBP/3QsBx548FrLn3zyT+y44yt63eaBB37NzjsHzc3NBjySJGnA/PGPf2Du3CfWutrm\nN7/5NZMmTebJJ//Ya4/14v6uiWHDhvXaL26sYf2+R0mSBtjixYvp6Ojol38WL15c0zF/+tM72XPP\n/wOsuSLlhc8efPDXADz22KOMG1dcirvPPvvx61/fz7JlSxk/fnyv+2xvf5qmJthhhwnPBym77bYH\nP/3pHQA88cQcbrzxG71s2fWi17vuuju//OX/AvCrX/2SPfYoAqLf/vZhInZ98R7KXYwatSUATz31\nFNdffy1Q3G++dOkz7LLL7tx/f7HPBQueoqlpGKNHNz9/DgDGjNmaFStW8MwzS5gw4eUA3HXXT1i1\nalWvY95tt9356U/vBGDRooVcddXlZZ0PEbFbr9t85zvf6tdvuyRJUjU9wRP8vp/+9wRPbPB4m2++\nOV/4wud56qn5QPGF1A03XM+MGe9eZ4/VW3+3664v9FS33DKb2277Xr+cD6/gkSRV2sSJk5k5c9Z6\n19lmm2YWLlz/UxF67nN9zj77X3jkkYe55JJPA7BixQr++Me5fOUrV9HU1MSoUaM588wPMn/+fE4/\n/QwANttsMyZOnNTjF//aV94sWPAURxzxz8Un5f3aM2YczoUXnsfMmcezevVqPvCBD/dSUdOLXr/v\nfSdy0UUXcOut32HEiBFcfPG/c8UVlzB58hRe9rKXvWgP8+Y9yUc/+mGeeWYJxx9/Mttttx2PP/4o\n73//cQwbNoyzzz6fsWO341e/+gWnnXYSq1at4l/+5aPPb3/hheezxRZb0Ny8FccddxKjRo3mc5/7\nFNtvvwMzZhzOxRdfyP/+7728/vV7Pz82gOnTD+L++3/JyScfy+rVXRx77An85je/5pZbZvOZz1za\n6/nfZ599GTVqdK+fqTGemjufB2/6DcOa6v/dYfPokSx7dkVdjzFhnx3ZZd8XN+mSpMFj4sTJtN2z\n4Sdf1Wprxm6wBxw3bjwf/ei5fOIT59LZuYq//nU5H/rQR7jjjh+ts8fqrb877LB/5BOfOJdTTz2R\n0aNHc+65n+yXMTR19bz5fxPU3r500x/kBowduxXt7UP3MnbHP7THD54Dx9+/4z/ssHfw9a/fxMiR\nLzzt4Fe/+iX/+Z83MWXKK9l66xYOOeSwtbZZsWIFp5xyApdcckVDwgn/DGzVyzPPqmew9jR3z/4J\nbzlpOqMY1ehS+sWNp3yL6R97S6PL6JV/l4f2+MFz4Pgd/1AeP6y/p/EWLUmS+qzpRZMjA2tdmdLd\nww8/xIknHsO73/0erzyRJElSXXiLliRJfXTzzbe8aNlrX7vn80+Y6mn33ffg2mu/We+yJEmSNIR5\nBY8kSZIkSVLFGfBIkiRJkiRVnAGPJEmSJElSxRnwSJIkSZIkVZwBjyRJkiRJUsUZ8EiSJEmSJFWc\nAY8kSZIkSVLFGfBIkiRJkiRVnAGPJEmSJElSxRnwSJIkSZIkVZwBjyRJkiRJUsUZ8EiSJEmSJFWc\nAY8kSZIkSVLFGfBIkiRJkiRVnAGPJEmSJElSxRnwSJIkSZIkVZwBjyRJkiRJUsUZ8EiSJEmSJFWc\nAY8kSZIkSVLFGfBIkiRJkiRVnAGPJEmSJElSxRnwSJIkSZIkVZwBjyRJkiRJUsVt1ugCpKGss7OT\ntrY5dT/OokXNLFy4rO7HmThxMsOHD6/7cSRJkiRJazPgkRqorW0Ol1/+OVpaWhpdykZbvHgxM2fO\nYsqUnRtdiiRJkiQNOQY8UoO1tLTQ2tra6DIkSZIkSRXmHDySJEmSJEkVZ8AjSZIkSZJUcQY8kiRJ\nkiRJFWfAI0mSJEmSVHEGPJIkSZIkSRVnwCNJkiRJklRxBjySJEmSJEkVZ8AjSZIkSZJUcQY8kiRJ\nkiRJFWfAI0mSJEmSVHEGPJIkSZIkSRVnwCNJkiRJklRxBjySJEmSJEkVZ8AjSZIkSZJUcQY8kiRJ\nkiRJFWfAI0mSJEmSVHEGPJIkSZIkSRVnwCNJkiRJklRxBjySJEmSJEkVZ8AjSZIkSZJUcQY8kiRJ\nkiRJFWfAI0mSJEmSVHGbNboASZKkTUlEjAa+BrwM2By4APgtcD3Fl2vzgaMyc2VEHAmcDnQCV2fm\nNY2pWpIkVZ1X8EiSJPWvo4HfZeZ04DDgEoqQ5wuZOQ14HDg2IkYB5wDTgQOAWRHR0piSJUlS1Rnw\nSJIk9a8OYNvy9TZAOzAN+G657FbgIGBv4L7MXJaZy4G7gX0HuFZJkrSJMOCRJEnqR5l5I7BTRDwK\n3Al8GBidmSvLVZ4GtgfGUYQ/a7SXyyVJkvrMOXgkSZL6UTmvztzMfGtE/A3w1R6rNK1j03Utf15H\nxzw6O1duaLUBt2RJR6NL6FfPPvsMCxbMbXQZvVqwoNEVNNZQHz94Dhx/oytorKE+foCxY/dY52cG\nPJIkSf1rX+AHAJn5YERsDzwbESMzcwUwAXgSmMfaV+xMAO5Z345bW3eoT8Ubaeut2xpdQr8aPXoM\n48bt1OgyejV27Fa0ty9tdBkNM9THD54Dx+/4h/L4N8RbtCRJkvrXY8AbACJiJ2Ap8EPg0PLzGcBt\nwH3AXhExJiKaganAXQNfriRJ2hR4BY8kSVL/ugq4JiLuBIYDJwIJfC0iTgDmAtdlZmdEnAncDqwG\nzstMv5aUJEkviQGPJElSP8rMZ4HDe/no4F7WnQ3MrntRkiRpk+ctWpIkSZIkSRVnwCNJkiRJklRx\nBjySJEmSJEkVZ8AjSZIkSZJUcQY8kiRJkiRJFWfAI0mSJEmSVHEGPJIkSZIkSRVnwCNJkiRJklRx\nBjySJEmSJEkVZ8AjSZIkSZJUcQY8kiRJkiRJFWfAI0mSJEmSVHEGPJIkSZIkSRVnwCNJkiRJklRx\nBjySJEmSJEkVZ8AjSZIkSZJUcQY8kiRJkiRJFWfAI0mSJEmSVHEGPJIkSZIkSRVnwCNJkiRJklRx\nBjySJEmSJEkVZ8AjSZIkSZJUcQY8kiRJkiRJFWfAI0mSJEmSVHEGPJIkSZIkSRVnwCNJkiRJklRx\nBjySJEmSJEkVt1k9dx4R04CbgYeAJuAB4GLgeopwaT5wVGaujIgjgdOBTuDqzLwmIjYDrgV2AlYB\nx2RmW0S8GvgisBp4IDNn1nMckiRJkiRJg9lAXMFzZ2ZOz8wDMvN04ALgssycBjwOHBsRo4BzgOnA\nAcCsiGgBjgAWZeb+wIXAReU+Pw+cWi5viYi3DMA4JEmSJEmSBqWBCHiaerx/E3Br+fpW4CBgb+C+\nzFyWmcuBu4H9gAOBb5fr/giYGhEjgEmZeX+3fby5fuVLkiRJkiQNbnW9Rau0W0R8B9iG4uqdUZm5\nsvzsaWB7YBzQ3m2b9p7LM7MrIrqA8cDCbuuu2YckSZIkSdKQVO+A51HgvMy8OSImA3f0OGbPq3tq\nWd7V4/N1rfu8jo55dHau3NBqm7QFCxpdQWMN1vF3dMxrdAn9qqNjHs3Nmze6jF4N1j8DA8XxN7qC\nxhvq52Ds2D0aXYIkSVJd1TXgycx5FJMsk5lzIuIpYK+IGJmZK4AJwJPAPNa+CmcCcE+5fDzwYDnh\nchPFxMzb9lh3vf8vubV1h/4ZUIWNHbsV7e1LG11GwwzW8S9b9lyjS+hXra07MG7cTo0uo1eD9c/A\nQHH8Q3v84DmQJEna1NV1Dp6IOCIizihfj6e45eqrwKHlKjOA24D7KIKfMRHRDEwF7gJ+CBxWrvsO\n4I7M7AQeiYip5fJDyn1IkiRJkiQNSfW+Reu7wDcj4p3ACOBE4DfA1yLiBGAucF1mdkbEmcDtFI8+\nPy8zl0bEjcBBEXEXsBw4utzvLOCqiGgC7s3MH9d5HJIkSZIkSYNWvW/RWkZx5U1PB/ey7mxgdo9l\nq4Fje1n3EeCN/VSmJEmSJElSpQ3EY9IlSZIkSZJURwY8kiRJkiRJFWfAI0mSJEmSVHEGPJIkSZIk\nSRVnwCNJkiRJklRxBjySJEmSJEkVZ8AjSZIkSZJUcQY8kiRJkiRJFWfAI0mSJEmSVHEGPJIkSZIk\nSRVnwCNJkiRJklRxBjySJEmSJEkVZ8AjSZIkSZJUcQY8kiRJkiRJFbdZowvQ0NXZ2Ulb25wBOdai\nRc0sXLis7seZOHEyw4cPr/txJEmSJEnqzoBHDdPWNofLL/8cLS0tjS6lXyxevJiZM2cxZcrOjS5F\nkiRJkjTEGPCooVpaWmhtbW10GZIkSZIkVZpz8EiSJEmSJFWcAY8kSZIkSVLFGfBIkiRJkiRVnAGP\nJEmSJElSxRnwSJIkSZIkVZwBjyRJkiRJUsUZ8EiSJEmSJFWcAY8kSZIkSVLFGfBIkiRJkiRVnAGP\nJEmSJElSxW3W6AIkSZI2JRFxLHAU0AU0AXsCuwHXU3y5Nh84KjNXRsSRwOlAJ3B1Zl7TmKolSVLV\neQWPJElSP8rMazLzgMycDpwLXAdcAFyWmdOAx4FjI2IUcA4wHTgAmBURLY2qW5IkVZsBjyRJUv18\nDPg48Cbg1nLZrcBBwN7AfZm5LDOXA3cD+zaiSEmSVH0GPJIkSXUQEXsBf8jMp4HRmbmy/OhpYHtg\nHNDebZP2crkkSVKfGfBIkiTVx3HAtb0sb1rH+utaLkmStEFOsixJklQfbwJOKV8vjYiRmbkCmAA8\nCcxj7St2JgD3rG+HHR3z6Oxcub5VGmLJko5Gl9Cvnn32GRYsmNvoMnq1YEGjK2isoT5+8Bw4/kZX\n0FhDffwAY8fusc7PDHgkSZL6WURsDyzNzFXloh8BM4Bvlv++DbgP+HJEjAFWA1Mpnqi1Tq2tO9St\n5o2x9dZtjS6hX40ePYZx43ZqdBm9Gjt2K9rblza6jIYZ6uMHz4Hjd/xDefwb4i1akiRJ/W97irl2\n1jgPODoifgK8DLiunFj5TOD28p/zMtOuVZIkvSRewSNJktTPMvN+4P92e/8UcHAv680GZg9gaZIk\naRPlFTx+iyelAAAgAElEQVSSJEmSJEkVZ8AjSZIkSZJUcQY8kiRJkiRJFWfAI0mSJEmSVHEGPJIk\nSZIkSRVnwCNJkiRJklRxBjySJEmSJEkVZ8AjSZIkSZJUcQY8kiRJkiRJFWfAI0mSJEmSVHEGPJIk\nSZIkSRVnwCNJkiRJklRxBjySJEmSJEkVZ8AjSZIkSZJUcQY8kiRJkiRJFWfAI0mSJEmSVHEGPJIk\nSZIkSRVnwCNJkiRJklRxBjySJEmSJEkVZ8AjSZIkSZJUcQY8kiRJkiRJFWfAI0mSJEmSVHEGPJIk\nSZIkSRVnwCNJkiRJklRxBjySJEmSJEkVZ8AjSZIkSZJUcQY8kiRJkiRJFbdZowuQNHR1dnbS1jZn\nQI61aFEzCxcuq+sxJk6czPDhw+t6DEmSJEnqjQGPpIZpa5vD5Zd/jpaWlkaXstEWL17MzJmzmDJl\n50aXIkmSJGkIMuCR1FAtLS20trY2ugxJkiRJqjTn4JEkSZIkSao4Ax5JkiRJkqSKM+CRJEmSJEmq\nOAMeSZIkSZKkijPgkSRJkiRJqjgDHkmSJEmSpIoz4JEkSZIkSao4Ax5JkiRJkqSKM+CRJEmSJEmq\nOAMeSZIkSZKkijPgkSRJkiRJqjgDHkmSJEmSpIoz4JEkSZIkSao4Ax5JkiRJkqSKM+CRJEmSJEmq\nOAMeSZIkSZKkijPgkSRJkiRJqjgDHkmSJEmSpIoz4JEkSZIkSao4Ax5JkiRJkqSKM+CRJEmSJEmq\nOAMeSZIkSZKkijPgkSRJkiRJqjgDHkmSJEmSpIoz4JEkSZIkSao4Ax5JkiRJkqSKM+CRJEmSJEmq\nOAMeSZIkSZKkitus0QVI0lDV2dlJW9ucATnWokXNLFy4rO7HmThxMsOHD6/7cSRJkiStzYBHkhqk\nrW0Ol1/+OVpaWhpdSr9YvHgxM2fOYsqUnRtdiiRJkjTkGPBIUgO1tLTQ2tra6DIkSZIkVZxz8EiS\nJEmSJFWcAY8kSZIkSVLFGfBIkiRJkiRVXN3n4ImILYCHgAuAHwPXUwRL84GjMnNlRBwJnA50Aldn\n5jURsRlwLbATsAo4JjPbIuLVwBeB1cADmTmz3mOQJEnqi7K3+TCwEvgY8CA19kANKlmSJFXcQFzB\ncw7w5/L1BcBlmTkNeBw4NiJGletMBw4AZkVEC3AEsCgz9wcuBC4q9/F54NRyeUtEvGUAxiBJklST\niNiGItSZCrwNeBd964EkSZL6rK4BT0QEsAvwPaAJmAbcWn58K3AQsDdwX2Yuy8zlwN3AfsCBwLfL\ndX8ETI2IEcCkzLy/2z7eXM8xSJIk9dGbgR9m5l8yc0Fmngi8idp6oH0bUbAkSaq+et+i9RlgJnB0\n+X50Zq4sXz8NbA+MA9q7bdPec3lmdkVEFzAeWNht3TX7kCRJGiwmAqMj4hagBTgfGNWHHkiSJKnP\n6hbwRMRRwM8yc25xIc+LNK1j0/Ut7+rx+brWXUtHxzw6O1dueMVN2IIFja7gxTo65jW6hH7X0TGP\n5ubN+7T+psTxD+3xQ9/PwUAZjP8NHGhD/RyMHbvHQB6uCdgG+AeKsOcOautfNtjXDNaeZsmSjkaX\n0K+effYZFiyY2+gyejXU/y4P9fGD58DxN7qCxhrq44f19zT1vILn/wKTIuLtwATgOWBZRIzMzBXl\nsieBeaz9bdUE4J5y+XjgwXLC5SaKSQm37bHuBv8fUmvrDhs/moobO3Yr2tuXNrqMtSxb9lyjS+h3\nra07MG7cTjWvv6mdA8c/tMcPfT8HA2Uw/jdwoHkOBtQCii+5VgNzImIpsLIPPdA6DdaeZuut2xpd\nQr8aPXrMoPxvGfh3eaiPHzwHjt/xD+Xxb0jd5uDJzH/MzL0zcx/gyxSTC/4IOLRcZQZwG3AfsFdE\njImIZooJCe8CfggcVq77DuCOzOwEHomIqeXyQ8p9SJIkDRa3A9MjoikitgWa6VsPJEmS1Gd1f0x6\nac0lx+cC10fECcBc4LrM7IyIMymaodXAeZm5NCJuBA6KiLuA5bwwj88s4KqIaALuzcwfD9AY+t3q\n1atZuvSZATnWiBGdLFlS36Rziy22ZOTIkXU9hiRJg11mzouIbwE/p7i9fCbwC2rsgRpVtyRJqrYB\nCXgy8/xubw/u5fPZwOwey1YDx/ay7iPAG/u7xka4666f8P3vf2dAQpGmpia6urrqeozx41/O+99/\nWl2PIUlSFWTm1cDVPRbX1ANJkiS9FAN1BY960dXVxXbbbceWW27Z6FL6xRZbePWOJEmSJEmNULc5\neCRJkiRJkjQwDHgkSZIkSZIqzoBHkiRJkiSp4gx4JEmSJEmSKs6AR5IkSZIkqeIMeCRJkiRJkirO\ngEeSJEmSJKniDHgkSZIkSZIqzoBHkiRJkiSp4gx4JEmSJEmSKs6AR5IkSZIkqeIMeCRJkiRJkirO\ngEeSJEmSJKniDHgkSZIkSZIqzoBHkiRJkiSp4gx4JEmSJEmSKs6AR5IkSZIkqeIMeCRJkiRJkirO\ngEeSJEmSJKniDHgkSZIkSZIqzoBHkiRJkiSp4gx4JEmSJEmSKs6AR5IkSZIkqeIMeCRJkiRJkirO\ngEeSJEmSJKniDHgkSZIkSZIqzoBHkiRJkiSp4gx4JEmSJEmSKs6AR5IkSZIkqeIMeCRJkiRJkirO\ngEeSJEmSJKniDHgkSZIkSZIqzoBHkiRJkiSp4gx4JEmSJEmSKs6AR5IkSZIkqeIMeCRJkiRJkirO\ngEeSJEmSJKniDHgkSZIkSZIqzoBHkiRJkiSp4gx4JEmSJEmSKs6AR5IkSZIkqeIMeCRJkiRJkirO\ngEeSJEmSJKniDHgkSZIkSZIqzoBHkiRJkiSp4gx4JEmSJEmSKs6AR5IkSZIkqeIMeCRJkiRJkirO\ngEeSJEmSJKniDHgkSZIkSZIqzoBHkiRJkiSp4gx4JEmSJEmSKs6AR5IkSZIkqeIMeCRJkiRJkirO\ngEeSJEmSJKniDHgkSZIkSZIqzoBHkiRJkiSp4gx4JEmSJEmSKs6AR5IkSZIkqeIMeCRJkiRJkipu\ns0YXIEmSJFVZZ2cnbW1zBuRYixY1s3DhsrofZ+LEyQwfPrzux5Ek9R8DHkmSJGkjtLXNYck+7Uxi\n0oAcbxua67r/J3iCtntgypSd63ocSVL/MuCRJEmSNtIkJvEqXtXoMvrNQup/lZAkqX8Z8EiSJPWj\niJgG3Aw8BDQBDwAXA9dTzH84HzgqM1dGxJHA6UAncHVmXtOYqiVJUtU5ybIkSVL/uzMzp2fmAZl5\nOnABcFlmTgMeB46NiFHAOcB04ABgVkS0NK5kSZJUZQY8kiRJ/a+px/s3AbeWr28FDgL2Bu7LzGWZ\nuRy4G9h3wCqUJEmbFG/RkiRJ6n+7RcR3gG0ort4ZlZkry8+eBrYHxgHt3bZpL5dLkiT1mQGPJElS\n/3oUOC8zb46IycAdrN1z9by6Z0PLn9fRMY/OzpUbWm3ALVnS0egS+tWzzz7DggVza16/o2Me22xC\nEyxDMabm5s0bXcaLLFjQ6Aoab6ifA8ff6Aoaa6iPH2Ds2D3W+ZkBjyRJUj/KzHkUkyyTmXMi4ilg\nr4gYmZkrgAnAk8A81r5iZwJwz/r23dq6Q32K3khbb93W6BL61ejRYxg3bqea11+27Lk6VtMYra07\n9OkcDJSxY7eivX1po8toqKF+Dhy/4x/K498Q5+CRJEnqRxFxREScUb4eT3Er1leBQ8tVZgC3AfdR\nBD9jIqIZmArc1YCSJUnSJsAreCRJkvrXd4FvRsQ7gRHAicBvgK9FxAnAXOC6zOyMiDOB24HVFLd1\n+bWkJEl6SQx4JEmS+lFmLgPe0ctHB/ey7mxgdt2LkiRJmzxv0ZIkSZIkSao4Ax5JkiRJkqSKM+CR\nJEmSJEmqOOfgkSRJkqSN0NnZSVvbnAE51qJFzSxcuKyux5g4cTLDhw+v6zEk9T8DHkmSJEnaCG1t\nc1iyTzuTmDQgx9uG5rrt+wmeoO0emDJl57odQ1J91BTwRMT2wCeBvYAu4OfA2ZnZXsfaJEmSGsb+\nR1JfTGISr+JVjS6jXyykvlcISaqPWufg+RLwS+A9wJHAI8BX6lWUJEnSIGD/I0mSKqPWW7RGZebl\n3d4/FBHvqEdBkiRJg4T9jyRJqoxar+AZXV6mDEBEvBzYoj4lSZIkDQr2P5IkqTJqvYLn48AvI+Ip\noAkYC7yvblVJkiQ1nv2PJEmqjJoCnsz8XkRMAV5FMcng7zNzeV0rkyRJaiD7H0mSVCU13aIVEbtR\nfGP1J+B04JqI2KWehUmSJDWS/Y8kSaqSWufguQ7YBfhviqdJfBe4ql5FSZIkDQL2P5IkqTJqDXiW\nZuYpwPLMvCIz/wNYXce6JEmSGs3+R5IkVUatkyxvGRGTga6ImEQx0aBPkZAkSZsy+x9JklQZtQY8\n44EfUTQ2Py6XddWlIkmSpMHB/keSJFVGrQHP2zPzobpWIkmSNLjY/0iSpMqodQ6eS+tahSRJ0uBj\n/yNJkiqj1it4doiIY3suzMxr+rkeSZKkwcL+R5IkVUatAc/WwP49lnUBNjiSJGlTZf8jSZIqo9aA\n53eZeUxdK5EkSRpc7H8kSVJl1DoHz5ciomnNm4ioNRiSJEmqKvsfSZJUGbUGPCuB73Z7f3dEHFqH\neiRJkgYL+x9JklQZtQY8HwTe2+39W4AP9X85kiRJg4b9jyRJqoxaA56mzFyy5k35urM+JUmSJA0K\n9j+SJKkyar2X/BcRcSNwJ0Uo9HfAL+tVlCRJ0iBg/yNJkiqj1oDnNOBIYG+Kx4N+A7ipXkVJkiQN\nAvY/kiSpMmoKeDKzKyJ+DSzLzO9EREtmrt7QdhGxJXAtMA4YCXwC+A1wPcU3YfOBozJzZUQcCZxO\ncenz1Zl5Tfm0imuBnYBVwDGZ2RYRrwa+CKwGHsjMmX0ZtCRJ0oa81P5HkiSpEWqagyciZgHXAOeX\ni86JiLNr2PTtwP9m5puAw4HPAhcAX8jMacDjwLERMQo4B5gOHADMiogW4AhgUWbuD1wIXFTu9/PA\nqeXyloh4Sy3jkCRJqtVG9D+SJEkDrtZJlt8DvAFYWL7/MPC2DW2UmTdl5qfLt68A/ghM44VHjt4K\nHERx6fN9mbksM5cDdwP7AQcC3y7X/REwNSJGAJMy8/5u+3hzjeOQJEmq1UvqfyRJkhqh1oBnafdL\nksvXNV+iHBH/A3wdmAWMzsyV5UdPA9tT3MLV3m2T9p7LM7OL4v738bzQaHXfhyRJUn/aqP5HkiRp\nINU6yfLjEXEu8LKIOITidqvf1nqQzNy3nDfnG0BTt4+a1rHJ+pZ31biP53V0zKOzc+WGVhtwixc/\n3egS+tWzzz7DggVza16/o2NeHatpjI6OeTQ3b96n9Tcljn9ojx/6fg4GyoIFja6g8Yb6ORg7do+X\nstlG9T+SJEkDqdaAZybFBMhPAu+luIXq8g1tFBGvA57OzD9l5gMRMRxYGhEjM3MFMKHc5zzWvgpn\nAnBPuXw88GA54XITxcTM2/ZYd73/L6m1dYeaBjnQWloep5iGaNMwevQYxo3bqeb1ly17ro7VNEZr\n6w5D+hw4/qE9fuj7ORgoY8duRXv70kaX0VCeg5fkJfU/kiRJjVDrLVqdFBMkvx04lGKS41ouiXkj\ncAZARIwDminm0jm0/HwGcBtwH7BXRIyJiGZgKnAX8EPgsHLddwB3ZGYn8EhETC2XH1LuQ5IkqT+9\n1P5HkiRpwNV6Bc8qiluj1lhzq9TwDWx3JfCViPgpsAVwMvBL4PqIOAGYC1yXmZ0RcSZwO8W97edl\n5tKIuBE4KCLuApYDR5f7nQVcFRFNwL2Z+eMaxyFJklSrl9r/SJIkDbiaAp7MHAYQEVsCF2fmKTVu\ntxw4spePDu5l3dnA7B7LVgPH9rLuIxRXB0mSJNXFS+1/JEmSGqHWW7QAyMy/ArvWqRZJkqRBx/5H\nkiRVQU1X8ETE9RSXJG8LrKhrRZIkSYOA/Y8kSaqSWufg+VH577/ghMaSJGlosP+RJEmVUWvA80S3\n16+NCAAy86f9XpEkSdLgYP8jSZIqo9aA5/vAH4F7KJ4gAcUlyzY4kiRpU2X/I0mSKqPWgCeAC4GR\nwL9m5p/qV5IkSdKgYP8jSZIqo9bHpD8J/HNEvA74UkT8jOJxoU44KEmSNkn2P5IkqUpqfYrWuRSX\nJAPcC8wA3gdMqlNdkiRJDWX/I0mSqqTWW7Taerz/fD/XIUmSNNi09Xhv/yNJkgatWgOeu+pahSRJ\n0uBj/yNJkiqj1oDnYWA+xRMkurr9e3Kd6pIkSWo0+x9JklQZtQY8P8/MA+paiSRJ0uBi/yNJkipj\nWI3rbRYRta4rSZK0KbD/kSRJlVHrFTzbAAsj4mngIeAe4NrMbK9bZZIkSY1l/yNJkiqjpoAnM3cH\niIjtgV2A/YHbgdfWrzRJkqTGsf+RJElVUusVPABk5nyKyQbviIi2ehQkSZI0mNj/SJKkKqgp4ImI\nXYHLgdcDq4GfAzPrWJckSVJD2f9IkqQqqXXiwMuAzwDjgQnAlcAX61WUJEnSIGD/I0mSKqPWW7Sa\nMvN73d5/OyJOq0dBkiRJg4T9jyRJqoxar+DZPCJet+ZNRLweGF6fkiRJkgYF+x9JklQZtV7B8yHg\nmxGxHdAEzAP+qW5VSZIkNZ79jyRJqoz1BjwRcRtwXGbeC+wSEVsDXcAhwH8B29e/REmSpIFj/yNJ\nkqpoQ7dofZXikaAnlO/HA98F3gf8fT0LkyRJahD7H0mSVDnrDXgy80Zgb2BqRPwP8P+AL2fm/pn5\nq4EoUJIkaSDZ/0iSpCra4CTLmbkwM48G2oB/zsyv17soSZKkRrL/kSRJVbOhOXj+SHHPOcAY4B0R\nsZJiosGuzNymzvVJkiQNKPsfSfr/7d19tF1nXSfwb5rSCEnbiIkprdCmlf4QO4wi2KG89IWXio44\nWhC1U4GCMloROi6ZqgsExnGQAUQU0FWtvPgKTEUKIwsqb6100QKO4ggPA2060mCaS5vSVNqGm8wf\nZ6fepvcmuek99+Q59/NZq6vn7vOcs3/7nHv2/eW7n70P0KMDfYvWE5alCgCAw4f+BwDozn4Dntba\njctVCADA4WCp+p+q+qYk/5DkVUk+nOQdGZ0e/5UkF7TWdlXV+UlenGQ2yaWttcuWYt0AwMpzwGvw\nAABwSF6W5KvD7Vcl+e3W2plJvpTkwqp60DDmnCRnJ7m4qtZPpFIAoHsCHgCAJVZVleQRSd6f0bV7\nzkxyxXD3FUmemtE3dV3bWtvZWrszydVJHj+BcgGAKSDgAQBYeq9L8p8zCneSZG1rbddw++YkD0my\nKcn2OY/ZPiwHAFi0A11kGQCARaiqC5J8orV242giz32smm/hfpbfY2Zma2Zndx1o2LK77baZSZew\npO6442vZtu3gL8U0M7M1D86pY6xo+c3MbM26dUdNuoz72LZt0hXMb9p+Bw7X9z85fH8Hlovtn3QF\nk7dx42kL3ifgAQBYWj+QZHNV/WCSE5LcnWRnVa1prd01LLspydbce8bOCUmu2d8Tb9hw/Hgqvp+O\nPXbLpEtYUmvXHpNNm0486PE7d949xmomY8OG4xf1GiyXjRuPzvbtt0+6jPuYtt+Bw/X9Tw7f34Hl\nYvtX9vYfiIAHAGAJtdZ+bO/tqnp5ki1JzkjyzCR/nOS8JB9Icm2S36+qY5LsHsa8eLnrBQCmg2vw\nAACMz97Trn41yXOq6mNJvjnJ24YLK1+S5IPDf69orTksCQAcEjN4AADGpLX2yjk/Pm2e+y9Pcvny\nVQQATCszeAAAAAA6J+ABAAAA6JyABwAAAKBzAh4AAACAzgl4AAAAADon4AEAAADonIAHAAAAoHMC\nHgAAAIDOCXgAAAAAOifgAQAAAOicgAcAAACgcwIeAAAAgM4JeAAAAAA6J+ABAAAA6JyABwAAAKBz\nAh4AAACAzgl4AAAAADon4AEAAADonIAHAAAAoHMCHgAAAIDOCXgAAAAAOifgAQAAAOicgAcAAACg\ncwIeAAAAgM4JeAAAAAA6J+ABAAAA6JyABwAAAKBzAh4AAACAzgl4AAAAADon4AEAAADonIAHAAAA\noHMCHgAAAIDOCXgAAAAAOifgAQAAAOjckZMuAAAAgH7Nzs5my5brl2Vdt966LrfcsnPs6znppJOz\nevXqsa8HlpKABwAAgEO2Zcv1ue1x27M5m5dlfQ/OurE+/w25IVuuSU455eFjXQ8sNQEPAAAA98vm\nbM6pOXXSZSyZWzL+WUKw1FyDBwAAAKBzAh4AAACAzgl4AAAAADon4AEAAADonIAHAAAAoHMCHgAA\nAIDOCXgAAAAAOifgAQAAAOicgAcAAACgcwIeAAAAgM4JeAAAAAA6J+ABAAAA6JyABwAAAKBzAh4A\nAACAzgl4AAAAADon4AEAAADonIAHAAAAoHMCHgAAAIDOCXgAAAAAOifgAQAAAOicgAcAAACgcwIe\nAAAAgM4JeAAAAAA6J+ABAAAA6JyABwAAAKBzAh4AAACAzgl4AAAAADp35LhXUFWvSfKEJKuTvDrJ\ndUnekVG49JUkF7TWdlXV+UlenGQ2yaWttcuq6sgkb01yYpJvJHlea21LVT0qyVuS7E7y9621i8a9\nHQAAB6OqHphR/7IpyZokv5bk73KQ/c9EigYAujfWGTxVdVaSR7bWzkjy9CRvSPKqJL/TWjszyZeS\nXFhVD0rysiTnJDk7ycVVtT7JTyS5tbX2xCS/nlFAlOF5XjQsX19V545zOwAAFuEHk1zXWjsrybOT\nvD6L638AABZt3KdofSzJs4bbO5KsTXJmkvcOy65I8tQkpye5trW2s7V2Z5KrM5r18+QkfzGMvTLJ\nGVX1gCSbW2ufmfMcTxnzdgAAHJTW2jtba68dfnxYkn/Kwfc/j1/uegGA6TDWgKe1tqe19vXhx+cn\neX+Sta21XcOym5M8JKMpzNvnPHT7vstba3uS7ElyXJJb5ozd+xwAAIeNqvqbJH+U5OIsrv8BAFi0\nsV+DJ0mq6oeSXJjkaUm+OOeuVQs8ZH/L9+xz/0Jj7zEzszWzs7sONGzZ7dhx86RLWFJ33PG1bNt2\n40GPn5nZOsZqJmNmZmvWrTtqUeOnie1f2dufLP41WC7btk26gslb6a/Bxo2nLfs6W2uPH64b+Mc5\nuN6l257mtttmJl3CkjqUnubBOXWMFS0/+/PFmbbfgUPpaaZp+xOfgcPVSt/+ZP89zXJcZPncJL+U\n5NzW2u1VdXtVrWmt3ZXkhCQ3Jdmaex+xOiHJNcPy45J8drjg8qqMLkz4LfuM3e+/kjZsOH6pNmdJ\nrV//pYxOw58Oa9cek02bTjzo8Tt33j3GaiZjw4bjV/RrYPtX9vYni38NlsvGjUdn+/bbJ13GRHkN\nlk9VPTrJza21L7fW/r6qVidZTP+zoMO1pzn22C2TLmFJ6Wnszxdr2n4H9DQ+A4erlb79BzLuiywf\nk+Q1Sf59a+22YfGVSc4bbp+X5ANJrk3ymKo6pqrWJTkjyVVJPpR/vYbPM5J8pLU2m+RzVXXGsPxH\nhucAADgcPCnJLyRJVW1Ksi6j/ueZw/0H6n8AABZt3DN4np3RbJt3VtXe06uek+QPquqFSW5M8rbW\n2mxVXZLkgxl99fkrhtk+f57kqVV1VZI7kzx3eN6Lk/ze8JyfbK19eMzbAQBwsH43o17n40m+KcnP\nJPl0kndU1U/nAP3PpIoGAPo21oCntXZpkkvnuetp84y9PMnl+yzbndG1e/Yd+7mMjo4BABxWhm/E\nOn+euw6q/wEAOBTj/pp0AAAAAMZMwAMAAADQOQEPAAAAQOcEPAAAAACdE/AAAAAAdE7AAwAAANA5\nAQ8AAABA5wQ8AAAAAJ0T8AAAAAB0TsADAAAA0DkBDwAAAEDnBDwAAAAAnRPwAAAAAHROwAMAAADQ\nOQEPAAAAQOcEPAAAAACdE/AAAAAAdE7AAwAAANA5AQ8AAABA5wQ8AAAAAJ0T8AAAAAB0TsADAAAA\n0DkBDwAAAEDnBDwAAAAAnRPwAAAAAHROwAMAAADQOQEPAAAAQOcEPAAAAACdE/AAAAAAdE7AAwAA\nANA5AQ8AAABA5wQ8AAAAAJ0T8AAAAAB0TsADAAAA0DkBDwAAAEDnBDwAAAAAnRPwAAAAAHROwAMA\nAADQOQEPAAAAQOcEPAAAAACdO3LSBQCwMs3OzmbLluuXZV233rout9yyc+zrOemkk7N69eqxrwcA\nAPYl4AFgIrZsuT5vetNvZv369ZMuZUns2LEjF110cU455eGTLgUAgBVIwAPAxKxfvz4bNmyYdBkA\nANA91+ABAAAA6JyABwAAAKBzAh4AAACAzgl4AAAAADon4AEAAADonIAHAAAAoHMCHgAAAIDOCXgA\nAAAAOifgAQAAAOicgAcAAACgcwIeAAAAgM4JeAAAAAA6J+ABAAAA6JyABwAAAKBzAh4AAACAzgl4\nAAAAADon4AEAAADonIAHAAAAoHMCHgAAAIDOCXgAAAAAOnfkpAsAAJg2VfWaJE9IsjrJq5Ncl+Qd\nGR1c+0qSC1pru6rq/CQvTjKb5NLW2mUTKhkA6JwZPAAAS6iqzkryyNbaGUmenuQNSV6V5Hdaa2cm\n+VKSC6vqQUleluScJGcnubiq1k+magCgdwIeAICl9bEkzxpu70iyNsmZSd47LLsiyVOTnJ7k2tba\nztbanUmuTvL4Za4VAJgSTtECAFhCrbU9Sb4+/Pj8JO9Pcm5rbdew7OYkD0myKcn2OQ/dPiwHAFg0\nAQ8AwBhU1Q8luTDJ05J8cc5dqxZ4yELLAQAOSMADALDEqurcJL+U0cyd26vq9qpa01q7K8kJSW5K\nsjX3nrFzQpJr9ve8MzNbMzu7a39DJuK222YmXcKSuuOOr2XbthsPevzMzNY8OKeOsaLlNzOzNevW\nHTXpMu5j27ZJVzC/afsdWOz7P23bn/gMHK5W+vYnycaNpy14n4AHAGAJVdUxSV6T5MmttduGxVcm\nOYv8JtgAABZ/SURBVC/Jnwz//0CSa5P8/jB+d5IzMvpGrQVt2HD8uMq+X449dsukS1hSa9cek02b\nTjzo8Tt33j3GaiZjw4bjF/UaLJeNG4/O9u23T7qM+5i234HFvv/Ttv2Jz8DhaqVv/4EIeAAAltaz\nk3xLkndW1aoke5I8J8kfVNULk9yY5G2ttdmquiTJBzMKeF7RWtO1AgCHRMADALCEWmuXJrl0nrue\nNs/Yy5NcPvaiAICp52vSAQAAADon4AEAAADonIAHAAAAoHMCHgAAAIDOCXgAAAAAOifgAQAAAOic\ngAcAAACgcwIeAAAAgM4JeAAAAAA6J+ABAAAA6JyABwAAAKBzAh4AAACAzgl4AAAAADon4AEAAADo\nnIAHAAAAoHMCHgAAAIDOCXgAAAAAOifgAQAAAOicgAcAAACgcwIeAAAAgM4JeAAAAAA6J+ABAAAA\n6JyABwAAAKBzAh4AAACAzgl4AAAAADon4AEAAADonIAHAAAAoHMCHgAAAIDOHTnuFVTVaUnek+T1\nrbU3V9W3JXlHRuHSV5Jc0FrbVVXnJ3lxktkkl7bWLquqI5O8NcmJSb6R5HmttS1V9agkb0myO8nf\nt9YuGvd2AAAAAByuxjqDp6oelOSNSa6cs/hVSX67tXZmki8luXAY97Ik5yQ5O8nFVbU+yU8kubW1\n9sQkv57k1cNzvCHJi4bl66vq3HFuBwAAAMDhbNynaN2Z5OkZzdTZ66wkVwy3r0jy1CSnJ7m2tbaz\ntXZnkquTPCHJk5P8xTD2yiRnVNUDkmxurX1mznM8ZZwbAQAAAHA4G2vA01rb3Vq7a5/Fa1tru4bb\nNyd5SJJNSbbPGbN93+WttT1J9iQ5Lsktc8bufQ4AAACAFWns1+A5gFWHsHzPPvcvNPYeMzNbMzu7\n60DDlt2OHTdPuoQldccdX8u2bTce9PiZma1jrGYyZma2Zt26oxY1fprY/pW9/cniXoOVvv3Lbdu2\nSVcwWRs3njbpEgAAxmoSAc/tVbVmmNlzQpKbkmzNvWfhnJDkmmH5cUk+O1xweVVGp3t9yz5j9/uv\nhA0bjl+66pfQ+vVfyugyRNNh7dpjsmnTiQc9fufOu8dYzWRs2HD8in4NbP/K3v5kca/BSt/+5bZx\n49HZvv32SZcBAMCYTOJr0q9Mct5w+7wkH0hybZLHVNUxVbUuyRlJrkryoSTPGsY+I8lHWmuzST5X\nVWcMy39keA4AAACAFWmsM3iq6tFJXpfR15zvqqpnJjk/yduq6oVJbkzyttbabFVdkuSDGX31+Sta\na7dX1Z8neWpVXZXRBZufOzz1xUl+r6pWJflka+3D49wOAAAAgMPZWAOe4Zuuzp7nrqfNM/byJJfv\ns2x3kgvnGfu5JE9aojIBAAAAujaJU7QAAAAAWEICHgAAAIDOCXgAAAAAOifgAQAAAOicgAcAAACg\ncwIeAAAAgM4JeAAAAAA6J+ABAAAA6JyABwAAAKBzAh4AAACAzgl4AAAAADon4AEAAADonIAHAAAA\noHMCHgAAAIDOCXgAAAAAOifgAQAAAOicgAcAAACgcwIeAAAAgM4JeAAAAAA6J+ABAAAA6JyABwAA\nAKBzAh4AAACAzgl4AAAAADp35KQLAICVanZ2Nlu2XL8s67r11nW55ZadY13HSSednNWrV491HQAA\nzE/AAwATsmXL9XnTm34z69evn3Qp99uOHTty0UUX55RTHj7pUgAAViQBDwBM0Pr167Nhw4ZJlwEA\nQOdcgwcAAACgc2bwAAAssao6Lcl7kry+tfbmqvq2JO/I6ODaV5Jc0FrbVVXnJ3lxktkkl7bWLptY\n0QBA18zgAQBYQlX1oCRvTHLlnMWvSvLbrbUzk3wpyYXDuJclOSfJ2Ukurqr+L8gEAEyEgAcAYGnd\nmeTpGc3U2eusJFcMt69I8tQkpye5trW2s7V2Z5Krkzx+GesEAKaIgAcAYAm11na31u7aZ/Ha1tqu\n4fbNSR6SZFOS7XPGbB+WAwAsmmvwAAAsr1WLXH6PmZmtmZ3ddaBhy+6222YmXcKSuuOOr2XbthsP\nevzMzNY8OKeOsaLlNzOzNevWHTXpMu5j27ZJVzC/afsdWOz7P23bn/gMHK5W+vYnycaNpy14n4AH\nAGD8bq+qNcPMnhOS3JRka+49Y+eEJNfs70k2bDh+fBXeD8ceu2XSJSyptWuPyaZNJx70+J077x5j\nNZOxYcPxi3oNlsvGjUdn+/bbJ13GfUzb78Bi3/9p2/7EZ+BwtdK3/0CcogUAMH5XJjlvuH1ekg8k\nuTbJY6rqmKpal+SMJFdNqD4AoHNm8AAALKGqenSS1yU5McmuqnpmkvOTvK2qXpjkxiRva63NVtUl\nST6YZHeSV7TWHJYEAA6JgAcAYAm11j6T0dee7+tp84y9PMnlYy8KAJh6TtECAAAA6JyABwAAAKBz\nAh4AAACAzgl4AAAAADon4AEAAADonIAHAAAAoHMCHgAAAIDOCXgAAAAAOifgAQAAAOicgAcAAACg\ncwIeAAAAgM4JeAAAAAA6J+ABAAAA6JyABwAAAKBzAh4AAACAzgl4AAAAADon4AEAAADonIAHAAAA\noHMCHgAAAIDOCXgAAAAAOifgAQAAAOicgAcAAACgcwIeAAAAgM4JeAAAAAA6J+ABAAAA6JyABwAA\nAKBzAh4AAACAzgl4AAAAADon4AEAAADonIAHAAAAoHMCHgAAAIDOCXgAAAAAOifgAQAAAOicgAcA\nAACgcwIeAAAAgM4JeAAAAAA6J+ABAAAA6NyRky4AAAAAejU7O5stW65flnXdeuu63HLLzrGv56ST\nTs7q1avHvh6WloAHAAAADtGWLdfntsdtz+ZsXpb1PTjrxvr8N+SGbLkmOeWUh491PSw9AQ8AAHC/\nLNcMBrMXOFxtzuacmlMnXcaSuSXj/5yx9AQ8AADA/bKcMxjMXgCYn4AHAAC436ZpBoPZC0CPfIsW\nAAAAQOcEPAAAAACdE/AAAAAAdE7AAwAAANA5AQ8AAABA5wQ8AAAAAJ0T8AAAAAB0TsADAAAA0DkB\nDwAAAEDnBDwAAAAAnRPwAAAAAHROwAMAAADQOQEPAAAAQOcEPAAAAACdE/AAAAAAdE7AAwAAANA5\nAQ8AAABA5wQ8AAAAAJ0T8AAAAAB0TsADAAAA0DkBDwAAAEDnBDwAAAAAnTty0gUcqqp6fZJ/l2R3\nkpe01j414ZIAABZNTwMALIUuZ/BU1ZOSfHtr7YwkL0jyxgmXBACwaHoaAGCp9DqD58lJ3pMkrbXP\nV9X6qlrXWts54boAABZDTwNA92ZnZ7Nly/VjX8+tt67LLbeM/0/kSSednNWrV499PUut14DnuCRz\npy/PDMu+OJlyDs1RRx2Vm2/enjVrjhr7uo444ojs3r17rOt46EM3L/oxO3bsGEMlk3Go2zItr4Ht\nX9nbnxzatqz07b8/jzvcTMt2TMBU9DRHPvABee/D3pc1R4y/pznyiNX5xu7Zsa7j60fduejH3JAb\nxlDJZNyQG3JsNh7S46aB7V/Z258c2muw0rd/y5br83ePuy4n5IQxVTUyM9ZnH7kpNyXXJKec8vBl\nWNvSWrVnz55J17BoVfV7Sd7XWrti+PmqJM9rrXXVDAEAK5ueBgBYKl1egyfJ1oyObu11fJKvTKgW\nAIBDpacBAJZErwHPB5M8M0mq6tFJbmqt3THZkgAAFk1PAwAsiS5P0UqSqvr1JGcmmU1yUWvtsxMu\nCQBg0fQ0AMBS6DbgAQAAAGCk11O0AAAAABgIeAAAAAA6d+SkCwCA5VJV35bke5P8Y2vt85OuBwDg\nUOhpmI+AZ4pU1ZlJ3pXkH4ZFq5I8IsnzW2v/a2KFLaOqOjHJZ5N8ali0JskvttY+Mbmqlk9VfXuS\nNyTZkGR1kk9ktP13T7SwZTJ8Bn6utfas4ecTk7y7tfbYyVa2POwDFv4MZLQveFuSdUn+Nsn/mVSN\n47TS94FMD/szn2c9jZ4m9gF6mhW8DzxUAp7p89HW2o/u/aGqLptkMRPy+dbaOUlSVU9M8vIk3zfZ\nksavqo5I8j8z+gaWq4dlv5XkZcN/K8W+V45faVeSX7H7gAN8BjYluaS19oUJlrhcVuQ+kKm0Yvdn\nc6zIz7Oe5h56mhW6D9DT3GNF7gPvDwHP9FuVZHNVXbc38a+qG5J8Z2vtXyZb2tismnN7U5Kbquoj\nSY5LsjbJF5O8qLU2bWn3U5N8bu8fgcFLk5xYVdcm+UKSU5Nc11q7aBIFTsjJVfXhJJuT7Ery5STP\naq19dbJlLZtVSV5ZVb+Y5PFJrk7y/tba6yZb1lgs9BnYk+QdSZ5fVadndDTwJa21b0ygxuUwdx94\nXJIvV9XDMjrad0SSG5M8p7W20v6hQP/0NHoaPY2eRk+jp9HT7IeLLK8cexa4PY2qqj5cVdckeV2S\n/9FaOzvJq5P8WWvtnClshJLRtNX/PXdBa+2uJHcleVSSl7bWvjfJY6vq30ygvuWwO/f+Q5Ak1w/J\n/1uTvGF4/1dKI7TXy4fPwI5h+6exEUoW/gwcm+TsjD7/Zw0///iyV7d85u4DX5vRfvC/JXlta+3M\nJFuTPGaSBcL9pKfR0+hp9DR6Gj2NnmYeZvCsHI8YEv9VGaWf02zuVL5K8q6q+q4J17Qc9mR0fu58\nvtBa2zrc/mSSyuic1mlzQ5LvqKqjhnP010+6oMPEqn3+P60W+gysSXJra+1vh5/fn+T0jI6ATaO5\n+8BTk7w7owM6P58krbVLJlgbLAU9zfTT0+hpFqKn0dPoafbDDJ6V4/NDyn12kn+edDHLpbXWknw9\nyUMnXcsy+HxGO/h7VNVRSY7Ovf9ArMqUHvFsrX05yZuTXFlVf53kNRMu6XAzle/7HAt9Bo7NaCr7\nXruXs6hJGs7P37sP9DefaaGnmX56Gj3NgUzl+z6HnmYfepqD44VZmaY98b5n+6rqwRkd3btpcuUs\nmw8leVhV/UByz8XZfiOj83VPrqpNw7LTk/zj5Mocr9bam1trT2qtPTnJT0+6nsPMtH/2F/oMPDvJ\nTFWdMIw7K8mnJ1Lh8th3H/iQjL6J5MnDsldW1TkTqg2W2rTv1/Q00dPoaeY17Z99Pc2InmaRBDzT\nb08WuAJ/VZ1bVS9c/pLG7tThXM2PJHlfRlefv8+Fx6rqjVX1bctf3ngMFxc7N8kLhwsQfjzJjoyu\nNv+FJL+e0dcrXt1a+9wUv/8Hpar+vKpWwmmq816roqq+v6peMIF6xmahz0Br7eVJfiXJnw2ndaxJ\n8qfDPxDeMrmKx2bffeDPJvnVJD81LDspyUdW+j6ALulp9DR6mnnoafQ0epqVvQ+Ya9WePdM+uw1W\ntqo6Mcm7937jCABAj/Q0APtnBg+sDJJcAGAa6GkAFmAGDwAAAEDnzOABAAAA6JyABwAAAKBzAh4A\nAACAzgl4AAAAADp35KQLAJZWVb0+yblJbk7yHUle2lp7+2SrAgBYHD0NwOKYwQPT58QkL2qtnZ3k\nQ5MuBgDgEOlpABbBDB6YPicm+fK+C6vqoUnenOSBSdYl+eXW2oer6g+TXNVau2wYtzujfcPLkqxu\nrb18oRVV1UeS/NfheU5McnVr7aFVtT7J7ybZkOTYJK9vrf3p8Jj/kOS3k3wxyXcneUZr7eNVdXaS\nvevaleSnWms3Do/5WpJPJVmV5Mw59d3QWnt7VT06ybVJTm6t/b859T0nyVNaaxfMrTfJR4b6Ksma\nJJ9srb2kqs5M8muttScO429I8uQk/5zk7Um+OcnRSd7dWnvNMOaZSd6Q5AtJHpLkutbaTw7b+dIk\nXx/qvWBubcNjvz3JpcN23Znkea21r1TVzyV51vC4zyf52dbaXVX1/CT/KcndST7aWvuVqtqY5A+H\n1/kbSS5KckeSt7bWzq6qVcNr95ettVdV1Q2ttc3D+t+T5JbW2oULvccAMEF6mn+tT0+jp4EDMoMH\nps/JSbbMs/wtSV7bWntKkh9K8gdVNd8+YM8hrnfVnMf+WpK/GtZ1ZpJXVdW3DPedmuRNw9G4zyZJ\nVT1wqO+Hh+W/k+R1w33fmuSrrbVzhvvmq++/J7lxgbrmG//NSf6utXZWa+1xSc6tqkfOM37v7W9N\n8hettScneUKSX66qdcN9Jye5tLV2TpJXz3ns+iQ/Ojzmr5L83Dx1/G6S32itnZXksiTPqqrHDq/D\nma21xye5LckLquphSX45yeOH5Q+pqocP2/7+oYF7eZIL9qn9uRk1wPfapqH5PHWemgDgcKGnOfD2\n6Gn0NHAPM3hgilTVcUm+1lq7e567z06yrqr2/pG8K6M/8kny0qr6jxk1NHP9ZFU9cVj+utbaFfM8\n7+uq6tbc+w/u2UkeU1XPnbOuzUm+muSRSS7f5zlOy+hI0eXD0Zkjkuwe7jslyf9dYJNTVT+c5DNJ\nHrDAkKdV1YeHbfi3w7IdSR5WVZ8YajsuoyNz/5LkwcPzPiD/uo+8OcmTqupnMzrStGYYt3Oo72Pz\nrHdbkrcPDeemJNfMM+b0vY9trb1zWO8vJjllTs0PGtb52CSf2vve7j1CVVWnZ2gcW2tXJblqOPKY\nqnpQkhck+a2hhr2v2aokvzr89/QFXjcAmBg9zbz0NHoa2C8BD0yXM5J8eoH77szoKMqtcxdWVZK8\nZs505tk5d7+9tfbyqvrOjKYAf2vu6xfmTGe+alh2V0ZTcD8zz/jTMzpqM9ddSW4cjhjt6xFJ/mGB\nbVqT5PlJvj/J+xYY88HW2k8mydBgJMmPJXlMRkeO9lTVdcPyTyf5u6r6eJKWUSOSJC9JctRwlClV\ntX2f+t40d4VVdWSSP0/yXa2166vqoiTfM09tu3PfmZR3JXlva+3n93nOH5lnbDI6ejXf8lVJ/ktG\njdLR+yx/bpL3ZtTkAcDhSE9zX3oaPQ3sl1O0YLr8WEZTZ+dz9XB/qmpDVf3mAuNWzXP7ltz3SNj+\nXJ3k2cO6HlhVb6qqI6rqEUl2t9a27jP+C0k2DE1XqupJVfWC4b4n5t5HiubW8bMZTSW+4yDr2vvY\nTUna0Ah9T0ZHrNa01na31s5vrT2ptfZTSW6fM/4fh9qekdGRvTXD0aSH577N2tFJZpPcWFXflNH0\n8TXz1POJJN83PO+PV9WvJfmbJE+vqrXD8p8Zjmhdl+R7906jrqp3VtV3D+P3PscTh+sPJMnxSR7d\nWtv3yOIDM5ry/FsH+ZoBwCToafZPT6OngfsQ8MCUqKo/S/IjSZ5fVR+u0cX3npLkkqp6VJIXJ/nh\n4UjO+5JcOTx03/O55/78zKq6PMm7krxontUudG77K5I8vKquSvLRjI4ibczoCMv6OfWdluQNrbU7\nM/oD/QdV9dEkr0zysaq6JKOLBv7lAuu8s7X2RweoZb5635XkjKGGH07y2iRvrKpjFxh/WZLnVdWV\nGV3w8Y+T/ElGR/fe0FrbPfdBwxHFP8noQoB/muQ1Sc6pqvP2ef4XJbloqON5Sd7SWvt0RkfPPjq8\nV2dmdG79P2X0uv51Vf1Nkutba3+b0TnqZ1fVxzK6TsBrh+fenOSSeV6DjRmdIz87z30AMHF6Gj1N\n9DRwSFbt2XOo1x4DDic1+naEc1prN+yz/Kokv9Ja+/hkKrunjhOT/OG+U5ZrzjcgAADoaQAOjRk8\nMD32ZP4px4d7inu41wcALC89DcAhMIMHAAAAoHNm8AAAAAB0TsADAAAA0DkBDwAAAEDnBDwAAAAA\nnRPwAAAAAHTu/wN93lXXv3VezwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1fbb4718d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "@use_result('hist_data')\n",
    "def task_hist_day_of_week(features_by_user, color_dict, usernames, **kwargs):\n",
    "    hist_params = []\n",
    "\n",
    "    for target, username in usernames.iteritems():\n",
    "        plot_data = features_by_user[target]['day_of_week']\n",
    "\n",
    "        hist_params.append(\n",
    "            HistParams(\n",
    "                data=plot_data, \n",
    "                title=u'Распределение часа начала сессии',\n",
    "                color=color_dict[target],\n",
    "                xticklabels=[u'Пн.', u'Вт.', u'Ср.', u'Чт.', u'Пт.', u'Сб.', u'Вс.'],\n",
    "                legend=username\n",
    "            )\n",
    "        )\n",
    "\n",
    "    color_hist(u'День недели начала сессии', u'Количество', *hist_params)\n",
    "    \n",
    "task.call('hist_day_of_week')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 7. Результаты анализа (закономерности) в исходных признаках"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Потенциально видны 2 полезных признака — это час начала сессии **(start_hour)** и день начала сессии **(day_of_week).**  Распределения этих признаков имеют моды (пики), сильно отличающиеся от мод на соответствующих графиках для других пользователей, т.е. данные признаки у сессий Элис чаще принимают значения, отличные от значений для сессий других пользователей (не-Элис).\n",
    "\n",
    "Распределения 2 других признаков (длина сессии **session_length** и число уникальных сайтов **#unique_sites**) cхожи для Элис и других пользователей. Отдельные значения у Элис отличаются, но это похоже на влияние выборки б**о**льшего размера (другие пользователи, не-Элис), задающей распределение, которое ближе к унимодальному, плавно спадающему. У Элис из-за м**е**ньшего размера данных распределение кое-где немного отклоняется от заданного (допустим, у Элис получилось сессий длины в 10 сайтов в целом больше, чем у большинства), но всё равно заметно, что распределение, которому соответствует этот признак, имеет похожую природу, что и для не-Элис."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 8. Новые признаки"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Для лучшей работы классификатора очень желательно извлечь из данных лишь о сайтах и времени посещения дополнительные признаки. К недостаткам математической модели, которой является классификатор, относится непонимание особенностей данных, присущее человеку. Поэтому необходимо сформировать новую информацию и, кроме всего двух признаков о названии и времени посещения сайтов, добавить дополнительные, которые переведут наши эмперические знания на язык математики.\n",
    "\n",
    "Кроме того, данные можно представить в виде сессии, т.е. рассматривать не по одной паре $site_i, time_i$, а брать, например, сразу 10 пар этих данных. И, допустим, сопоставлять эти 10 пар из конкретной сессии одному пользователю. Так данных в строке будет больше.\n",
    "\n",
    "Можно пробовать разные длины сессий, делать сессии перекрывающимися.\n",
    "\n",
    "Ниже представлены новые признаки, которые могут быть получены из исходных данных.\n",
    "\n",
    "* **time_diff1, time_diff2, ..., time_diff10** – интервал между двумя сайтами в секундах, время пребывание на сайте;\n",
    "* **session_timespan** – количество секунд между началом и концом сессии;\n",
    "* **unique_sites** – количество уникальных сайтов;\n",
    "* **start_hour** – час начала сессии;\n",
    "* **day_of_week*** – день недели начала сессии;\n",
    "* **day, month, year** – день, месяц и год;\n",
    "* **through_hour** – сквозной час (сколько часов прошло с 1 января 1970 года, т.е. с \"эпохи\");\n",
    "* **second_of_day** – номер секунды дня;\n",
    "* **avg_seconds_on_top_sites** – среднее количество секунд, проведённое на топ-10 сайтах в текущей сессии;\n",
    "* **top_site_ind** – индикаторы топовых сайтов, всего 10 (1 – если в данной сессии был заход на соотв. топовый сайт, иначе 0);\n",
    "* **facebook_num** – сколько раз пользователь зашёл на facebook.com;\n",
    "* **www_sites** – сколько раз пользователь зашёл на сайт, начинающийся с префикса www.;\n",
    "* **small_time_diffs** – сколько в текущей сессии встречается коротких сессий, меньших 40, 20, 10, 5, 3, 1 секунды\n",
    "* **num_next_changes** – сколько раз следующий сайт в сессии отличался от предыдущего (отражает, насколько пользователь склонен менять сайт при очередном перемещении);\n",
    "* **num_next_same** – количество пар, когда предыдущий сайт такой же, как последующий; пара считается законченной, если был найден повторяющийся сайт, т.е. следующий сайт, если он такой же - не принадлежит к найденной паре;\n",
    "* **num_most_freq_site** – сколько раз был посещён самый частый сайт в данной сессии; отражает, склонен ли пользователь часто ходить по одному сайту;\n",
    "* **avg_num_dots, min_num_dots, max_num_dots** – среднеарифметическое, минимальное и максимальное количества точек (разделителей) в имени сайта в сессии;\n",
    "* **min_time_diff** – минимальный интервал между временем посещения двух сайтов в сессии.\n",
    "* **start_0_6** – началась ли сессия в интервале [0, 6) часов.\n",
    "* **start_6_12*** – началась ли сессия в интервале [6, 12) часов.\n",
    "* **start_12_18** – началась ли сессия в интервале [12, 18) часов.\n",
    "* **start_18_24** – началась ли сессия в интервале [18, 24) часов.\n",
    "\n",
    "Признаки, помеченные звёздочками (т.е. день недели и индикатор, началась ли сессия пользователя от 6 утра и до 12 дня) дали улучшение модели и были добавлены в новую модель.\n",
    "\n",
    "Кроме того, над признаками названий сайтов лучше поработать. Во-первых, нужно их перевести в числовой вид. Большинство моделей успешно работают только с числами. Во-вторых, после перевода сайтов в их номера (числа) возникнет проблема, что признак $sitenum_k$ будет зависеть от того, как мы разбили сессию. Т.е. в строке, допустим 2 на месте сайта 3 будет один номер, в следующей строке на этом же месте вероятнее всего окажется другой номер, никак не связанный по смыслу с предыдущим номером сайта. Признак будет иметь нечёткий смысл, будет сильно меняться и модель будет обучаться хуже.\n",
    "\n",
    "Если же применить подход **bag-of-words**, т.е. на месте $site_i$ вписывать не номер i-го сайта в сессии, а, например, количество, сколько раз встретился $site_i$ в сессии, то признак сразу приобретает чёткий смысл для всех строк. Т.е. новые строки будут давать дополнительную информацию: классификатор сможет настраивать корректно веса модели исходя из конкретного признака $i$.\n",
    "\n",
    "Будем использовать в качестве новых признаков также указанные количества использованных сайтов, а номера сайтов исключим из модели, т.к. новые признаки о количестве сайтов уже содержат информацию о посещённых сайтах. Последовательность перехода между сайтами, которая не учтена в данном преобразовании, можно учесть в дополнительных признаках, например, в таких признаках как **num_next_same** (см. описание признаков выше)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 9. Отделимость классов"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Строить картинку отделимости классов методом **t-SNE (t-Distributed Stochastic Neighbor Embedding),** например, затратно с точки зрения процессорного времени. Так как у нас есть другие более быстрые и не менее эффективные методы визуального анализа, приведу лишь сгенерированную ранее картинку **t-SNE** для разных классов (когда была поставлена другая задача на тех же данных: отделить всех пользователей).\n",
    "\n",
    "Видно, что большое количество пользователей отделить очень сложно. Тут, наверное, ошибался бы даже человек. Привожу картинку с целью дать представление о сложности такой задачи. Классы, окрашенные в разные цвета, расположены при проекции в 2d-пространство без каких-то понятных законов, они особенно нигде не кластеризуются."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": false
   },
   "source": [
    "<img src='https://www.dropbox.com/s/1hho0ywh5bafxmj/tsne_many_users.png?raw=1'>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 10. Обучение модели, кросс-валидация и кривые валидации"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 10.1 Обучение модели и кросс-валидация"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "collapsed": true
   },
   "source": [
    "Новые признаки в модель добавлялись постепенно, по одному. При этом я использовал три метрики для оценки бинарной классификации (Элис, не-Элис): **accuracy, precision и recall.** При этом я сравнивал предыдущую модель с новой, в которую добавлялся новый признак.\n",
    "\n",
    "Об однозначном улучшении модели бинарной классификации (для любой бизнес-задачи) можно сказать более или менее однозначно, если **accuracy, precision и recall** улучшились. Почему именно все три эти метрики для любой задачи?\n",
    "\n",
    "1. Если у нас улучшилась только **accuracy,** значит, мы стали точнее различать положительные и отрицательные классы. В целом это неплохо, но если ухудшилась метрика **recall,** значит, теоретически бизнес-задача может быть решена хуже, чем с предыдущей моделью, с худшим параметром **accuracy.** Если в бизнес-задаче стоит условие, чтобы находить как можно больше сессий Элис, то новая модель работает с точки зрения клиента хуже. Представим, что в выборке у нас очень много объектов отрицательного класса (не-Элис). Улучшение **accuracy** может означать, что модель стала давать меньше положительных срабатываний и стала лучше находить объекты отрицательного класса. В то же нельзя однозначно сказать что-то о качестве нахождения классов Элис, ведь выборка сильно несбалансирована.\n",
    "\n",
    "2. Улучшение **recall,** с другой стороны, означает нахождение большего числа объектов Элис. Но если в нашей бизнес-задаче стоит условие, чтобы не находить положительные объекты попусту (допустим, чтобы не посылать руководителю лишних логов), а найти только точно соответствующие Элис, то с точки зрения этой задачи улучшение **recall** не обязательно означает улучшение модели.\n",
    "\n",
    "3. И аналогично для **precision.** Улучшение **precision** лучше подойдёт, если клиент хочет точнее найти объекты Элис, но при этом многие объекты могут быть не обнаружены. Так, если бы мы строили классификатор для определения потенциальных сотрудников (должен или не должен быть нашим сотрудником) или для определения потенциальных клиентов, которым стоило бы отправить сообщение о новогодней акции, то мы рисковали бы многих их них проигнорировать.\n",
    "\n",
    "Так, в связи с тем, что задача недоопределена (что именно хочет получить заказчик), будем смотреть на все три эти метрики. В качестве примера я привожу 2 задачи, одна из которых содержит дополнительный добавленный параметр, а другая — два дополнительных параметра. При этом модели сравниваются по указанным метрикам: **accuracy, precision, recall.**\n",
    "\n",
    "**Кросс-валидация** для усреднения данных метрик проводилась на 3 фолда."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.9912447\n",
      "Precision: 0.6634272\n",
      "Recall: 0.0679164\n",
      "Roc_auc: 0.9208248\n"
     ]
    }
   ],
   "source": [
    "@use_result('my_load_data')\n",
    "def task_my_all_model_quality_prev(train_features, train_targets, train_site_cnts, **kwargs):\n",
    "    sgd_logit = SGDClassifier(loss='log', random_state=17, n_jobs=-1)\n",
    "\n",
    "    trunc_train_features = train_features[['day_of_week']]\n",
    "\n",
    "    X = sparse_hstack( (csr_matrix(trunc_train_features.values), train_site_cnts) )\n",
    "\n",
    "    y = train_targets\n",
    "\n",
    "    cross_val_score_part = partial(cross_val_score, estimator=sgd_logit, X=X, y=y, cv=3)\n",
    "\n",
    "    print('Accuracy: %.7f' % cross_val_score_part(scoring='accuracy').mean())\n",
    "    print('Precision: %.7f' % cross_val_score_part(scoring='precision').mean())\n",
    "    print('Recall: %.7f' % cross_val_score_part(scoring='recall').mean())\n",
    "    print('Roc_auc: %.7f' % cross_val_score_part(scoring='roc_auc').mean())\n",
    "    \n",
    "task.call('my_all_model_quality_prev')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Признак day_of_week дал улучшение параметров и был добавлен к модели ранее. Попробуем также добавить сгенерированный признак **start_6_12,** означающий, началась ли сессия в интервале [6; 12) часов."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.9914616\n",
      "Precision: 0.6942328\n",
      "Recall: 0.1027390\n",
      "Roc_auc: 0.9508164\n"
     ]
    }
   ],
   "source": [
    "@use_result('my_load_data')\n",
    "def task_my_all_model_quality(train_features, train_targets, train_site_cnts, **kwargs):\n",
    "    sgd_logit = SGDClassifier(loss='log', random_state=17, n_jobs=-1)\n",
    "\n",
    "    check_feature = 'start_6_12'\n",
    "\n",
    "    trunc_train_features = train_features[['day_of_week', check_feature]]\n",
    "\n",
    "    X = sparse_hstack( (csr_matrix(trunc_train_features.values), train_site_cnts) )\n",
    "\n",
    "    y = train_targets\n",
    "\n",
    "    cross_val_score_part = partial(cross_val_score, estimator=sgd_logit, X=X, y=y, cv=3)\n",
    "\n",
    "    print('Accuracy: %.7f' % cross_val_score_part(scoring='accuracy').mean())\n",
    "    print('Precision: %.7f' % cross_val_score_part(scoring='precision').mean())\n",
    "    print('Recall: %.7f' % cross_val_score_part(scoring='recall').mean())\n",
    "    print('Roc_auc: %.7f' % cross_val_score_part(scoring='roc_auc').mean())\n",
    "    \n",
    "task.call('my_all_model_quality')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Видим, что все 3 метрики **(accuracy, precision, recall)** на кросс-валидации одназначно улучшились. Добавим также признак **start_6_12** к итоговой модели."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Примечание.** Попытка настройки гиперпараметра $ \\alpha $ модели SGDClassifier с логистической функцией потерь (log_loss) в данном случае *не дало* сколько-нибудь улучшения в качестве классификации. На тренировочный данных был найден гиперпараметр, близкий по значению к $alpha$ по умолчанию, что дало незначитальное улучшение, а классификация на тестовых данных (которые были посланы на **kaggle.com)** стала с настроенным параметром даже хуже, метрика **kaggle.com** ухудшилась."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Результат.** Оба бенчмарка: \"sgd_logit_benchmark.csv\" и \"Logit +3 features\" побиты. См. [результаты](https://www.kaggle.com/c/catch-me-if-you-can-intruder-detection-through-webpage-session-tracking2/leaderboard) на Leaderboard для команды **[YDF & MIPT] Coursera_zaharov** (score=0.94069). Очки для бенчмарка \"Logit +3 features\" score=0.92784. В курсе Яндекса/МФТИ по анализу данных на coursera.com требовалось побить эти уровни для получения максимальной оценки за работу."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 10.2. Кривые валидации"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Кривые валидации позволяют определить зависимость какой-либо метрики качества от гиперпараметра модели. Будем использовать быстрый классификатор градиентного спуска с функцией потерь log_loss (логистическая функция потерь).\n",
    "\n",
    "Исследуем гиперпараметр модели $\\alpha$ (*alpha*). Напомню, что этот гиперпараметр влияет на чувствительность модели (связан с регуляризацией). При понижении чувствительности модель будет меньше реагировать на отклонения в данных и с меньшей вероятностью переобучится, с другой стороны, модель может упускать некоторые закономерности в данных. Следует подбирать оптимальный параметр чувствительности модели.\n",
    "\n",
    "Построим кривую для двух метрики **ROC_AUC**."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Кривая для метрики *roc_auc.***"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYwAAAESCAYAAADuVeJ5AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzt3XmcHGWd+PFPVV9zZmaSDLlJOL8xiCxBBFkuRQWUBVcE\nAeVS1F0QZXVdfrLiLrjgLi6YFXFdQQ5dVl0REBdRDIci4hFQbr5cSQg5yEySmcw90931+6OqJz1X\nT/Vkurtm5vt+Gbu7urqfZ5hvzbeeeup5HsfzPIwxxpjxuJWugDHGmKnBEoYxxphQLGEYY4wJxRKG\nMcaYUCxhGGOMCcUShjHGmFDipS5ARK4DDgeywCWquibYvhC4HfAAB9gbuFRVfyAiHwY+DwwAX1LV\n+0pdT2OKYXFtZiKnlOMwRORo4O9V9WQRWQ7crKpHjLJfDHgIOAGoAh4DDgbqgStU9W9KVkljimRx\nbWaqUl+SOg64G0BVXwAaRaRulP3OA36sqt3Au4Bfqmq3qr5hB5WJIItrMyOVOmHMB1ryXrcG24a7\nAPhO8HwZUCsiPxGRX4nIO0tbRWOKZnFtZqSS92EM4wzfICKHA8+ramfePrOB9wN74Tfpl5athsYU\nz+LazAilThibGHrmtRDYPGyfk4DVea/fAH6rqh7wqoh0iMhcVW0dqxDP8zzHGXHMGjNZhgeXxbWZ\nLooKsFInjPuBfwZuFJGVwEZV7Rq2z6HA94d95hYRuQb/jKy20EEF4DgOLS0dk1frkJqb6ytSbiXL\nnqk/8zDTOq7B4msmlJsruxgl7cNQ1ceAx0XkUWAVcJGInCsip+TtNh/YmveZTcAdwO+Ae4FPlbKO\nxhTL4trMVCW9rbaMvJl4ZmA/c9nKrdR1oYrENVh8zYRyg7KLim0b6W2MMSYUSxjGGGNCsYRhjDEm\nFEsYxhhjQrGEYYwxJhRLGMYYY0KxhGFMRE2TW97NNGIJwxhjTCiWMIwxxoRiCcMYY0woljCMMcaE\nYgnDGGNMKJYwjDHGhGIJwxhjTCiWMIyJKBuGYaLGEoYxxphQLGEYY4wJxRKGMcaYUCxhGGOMCcUS\nhjHGmFAsYRhjjAnFEoYxxphQLGEYE1E2DMNEjSUMY4wxoVjCMMYYE4olDGOMMaFYwjDGGBOKJQxj\nospmHzQREy91ASJyHXA4kAUuUdU1wfaFwO34N4M4wN7Apar6g+D9KuAZ4EpV/W6hMtY8/wZL59aU\n7ocwZphyxLUxUVPSFoaIHA3sq6pHABcAX8+9p6qbVPUdqvpO4F3AeuCevI9fDmwLU87NP31m8ipt\nzDjKFdfGRE2pL0kdB9wNoKovAI0iUjfKfucBP1bVbgAREWA5cG+YQtIZa7qbsipLXF/+7cdYu3nn\npFTYmMlQ6oQxH2jJe90abBvuAuA7ea+vBT6L36Q3JmrKEtdPv9zKc+u2T7SOxky6cnd6jzhQRORw\n4HlV7Qxenw38VlXXj/WZEayBYSqrNHFtTMSUutN7E0PPvBYCm4ftcxKwOu/1+4C9ROSvgMVAr4hs\nUNUHCxXU3Fw/CdUtXqXKrWTZM/FnHqZscV1bm5pxv+eZGF8RietxlTph3A/8M3CjiKwENqpq17B9\nDgW+n3uhqmfknovIPwFrxzuoPDxaWjomrdJhNTfXV6TcSpY9U3/mYcoS1wCdnX0z6vc8U+Orkj9z\nMUp6SUpVHwMeF5FHgVXARSJyroickrfbfGDr7pRjt6ubcipXXINdbTXRUvJxGKp62bBNTw97/6AC\nn72iJJUyZjdZXJuZaFqM9LazMDNtWfPZRMi0SBh2UJnpyiLbRMn0SBjGGGNKblokDGtgmGnLYttE\nyPRIGJWugDElYrFtomRaJAxjjDGlNz0Shl2TMtOUZ7FtImR6JAxjjDElNy0Shp2DGWNM6U2PhGEZ\nw0xTFtsmSqZFwjDGGFN60yJheHZRykxTFtkmSkJNPhgsPzk7eJkCblfVt5WsVsWyo8pMWxbcJjrG\nbWGIyD8ArwMKPA78KfgXGXad1xhjSi9MC+ODwB7AL1T1HSJyMrC0tNUqjl2SMhNx2mknM3SlVA9w\n+NGPflKhGo1kJ0MmSsL0YXSoaj+QBFDVe4BTCn/EmOhraprNqlU38LWvfYOmpiZWrfomq1bdUOlq\nGRNZYVoYO0Tkw8AzInIL8Bz+GsbRYWdhZgJSqRSLFi0GoL+/n6qqKubMmVvhWg1lLQwTJWFaGOcA\njwJ/B7yEv4D9GQU/UWZ2TJmJ6O3t5ec/v5e7774Dz8vy+c9fwlNP/bnS1RrCLreaKAmTMK4DulW1\nW1WvBv4fcEFpq1UcOwszE/G5z13KI488zJNP/pkrr/xX/vVfr+Ub31hV6WoNZbFtIiTMJak/AqtF\n5EfAr4EbgO+XtFZFs6PKFG/58hVcddVXh2y7/vr/qlBtRmeRbaJk3IShqt8RkduAi/ETxXtU9ZmS\n16wI1sIwE3HaaSfjOE4QP7kgitZdUpYxTJSMmzBE5BFy9xtCDfBNEUFVjy515cKyY8pMxGmnncGD\nD67m/PM/zmGHvb3S1RmV9WGYKAlzSeqLJa/F7rImhpmA008/ixNOOIlbbrmRe+65kwsv/MzgXVNR\nYaFtoiRMp/cfgG2q+itgCSDBtsiwY8pMxJYtW+ju7uZDH/owp5xyKldc8UW+9a1vVLpaxkRWmBbG\nXcBCEXkKSAA7gBOAD5SyYsWwszAzEVdd9U9DXqdSKZ57LlLdcxbbJlLCJIxaVX2LiDyvqm8CEJFf\nlbheRbHrvGYiRrsj6pVXXubPf34CgL/4i5XlrtIIFtsmSsIkjIHgcXPetkwJ6jJxdkyZCbjppm+N\n2LZ69S94y1v+AohGwrDYNlESJmEcISKvAXsEjw4QrfkTjJkA1x2tC8/hssv+aZTtlWH5wkRJmIQh\nJa/FbrLrvGYiPvrRT4zY9uyz0erDsIxhoiRMwngdOAs4FD98f6eqoUd6i8h1wOFAFrhEVdcE2xcC\nt7NrjMfewKWq+gMRuQY4EogB/6qqdxUqw44pMxHZbHbCny1HXIP1YZhoCZMwvo6/HsbD+AfA6SJy\nuKp+ZrwPisjRwL6qeoSILAduBo4AUNVNwDuC/WLAQ8A9InIssCL4zGz8xZoKH1jWxDATcMwxhwUj\nvXfFj+M4BT7hK1tcYydDJlrCJIw3q+oxea+/EYz+DuM44G4AVX1BRBpFpE5VO4ftdx7wY1XtDu7A\n+n2wvQ2oERFHVcc8duygMhPxyCN/HLFtw4bXwny0LHENWHCbSAmTMJIi4qpqFgbPmkKtBQ7MB9bk\nvW4Ntr08bL8LgHcDBAdQT972n413UFkDw0xEJpPhD394jLa2NgAGBgb47ndv5o47fjreR8sS12D5\nwkRLmD/89wJ/zBt78Q7gBxMsb0R7X0QOB54ffnYmIqcA5wPvGe9LPc+jubl+glXaPZUqt5JlT5ef\n+bOf/Szt7e2oKitXruTJJ5/k4osvnkgZJYlrgKqqxIz7PU+X+JoK5RYrzGy1/yIiq4HD8E94Pqmq\nYacG2YR/5pWzkKHjOQBOAlbnbxCR44EvAMerasd4hXhAS8u4u0265ub6ipRbybKn08+8YcNGvvnN\nm/jUpz7B5ZdfxZYtm/nv/76VY445fkS5w5QlrgG6u/tn1O95OsVX1MvNlV2MceeSCi5BzQVSwb8m\nERm/Z9B3P/DB4HtWAhtVtWvYPocCT+aVNwu4BjhJVdvDFOLZNSmzGzKZDH19fcyfv4C1a18N85Gy\nxLXPYttER5hLUrcAS4Hf4je9v4h/sHx8vA+q6mMi8riIPIo/OvwiETkXaFPV3KID84GteR/7EDAH\n+N8gMXnAOar6+ljlWL4wE3HIIYdy++23cdRRx/Kxj53NggULQt1qW664BottEy1hEsZyVX1b7kUQ\n7L8LW4CqXjZs09PD3j9o2OsbgRvDfn+O53mhbok0JudjH/skmUyGWCzGgQe+he3bt4deF6N8cV3s\nJ4wpnTAJY6OIVKlqb/A6BYRqt5dTbpSUMWFt2bJl8Hlz8zyam+dx5ZWX89JLSlRW3rOBeyZKwiQM\nB3glaH67+J3fz4jIdwFU9ZwS1i80z/PAWhimCGed9QEaG5uGbGtra+PBBx+tUI1GshaGiZKw62Hk\nj0gd9yb1SrADyxTrgAMOHDHF+cUXf7JCtRmd3dBhoiRMwnhotI2qGmpIbLnYcWWK1dKylQceuJ+l\nS/di7733wXXdyPWDZS2uTYSESRivAuuDf7mjyQPeWapKTYwdWaY4H/jAafzpT09w1113sH79Ot70\nphVs2TJ8OEWFWVibCAmTMN4GfAV4AviXUe43jwRrYZhinX76WYPP0+k0a9b8gWuuuYqrr74CIBLr\nYlint4mSMCO9nwCOF5G/Bn4mIjeq6n+XvmrFsYRhdkc8Hufww4/g3/7ta3R2VmbU7WisD8NEybgJ\nQ0Ty74L6b+BzInKRqoa7Yb1M7EzMFOu0004OpjeHXdd+onE7bY7lCxMlYS5J7TXs9d2lqMjusgPL\nFOu0087gwQdXc/75Hw89YK/crNPbREmYhPHlktdiEljT3RTr9NPP4oQTTuKWW27knnvu5MILP8Oi\nRYsrXa2hLK5NhIRJGGmG3quRmwcnVpIaTZAdVqZYuZHeH/rQh3nttXVcccUXWbnyrfzN33yqwjXb\nZeKLyBoz+cJ0ersAIvKQqr6j9FWaGDsRM8W66qqhd0GlUimeffbpMfauDGs5mygJu3IeRPwkPmsH\nlinS8FHeUWRhbaIkzF1SHw2eLsx7jqreXLJaTYAdWKZYF100+gz9N9xQ9KSyJZO1Xm8TIWFaGEcF\nj4/lPfeAiCUMO7BMcbq7u/j0pz9X6WoUZHFtoiRMH8b5IuICe6jqlvH2rxQ7rkyx6utncfDBh1S6\nGgVZA8NESZhLUu8EvgP0ActF5GvAalW9t9SVK4adiZlivfLKS1x99RUsXryEvffehwMOOJCmptmV\nrtYQFtcmSsZd0xu4GjicXYvcXwVcXrIaTZB1eptiXXvt9Rx22BE4jsPDDz/I3/7tx7juun+rdLWG\nsLA2URKmD6NTVd8QEQBUtVVE+ktbreLZgWWKtXz5CpYvXzH42vM8vvCFaPVp2ImQiZIwCaNHRI4B\nHBFpAs4Aesf5TNnZgWUmYu3aV2lvbwNgYKCfDRsitcyLxbWJlDAJ40LgP4FDgVeAR4BPlLJSE2HH\nlSnWqlX/zh//+Du2bdvG4sVL2Ljxdc488yOVrtYgx7G4NtES5i6pDcBJZajLbrHOQVOs559/lttv\nv4OLL/4k11//X7zwwvP8+tejLjBZEf5MuhbXJjrC3CV1DHAd8Cb88RdPAZ9V1cdKXLei2O2HpliJ\nRAKAgYEBPM9j+fI3ccMNqypcq11cxyFrk0mZCAlzSWoV8DngUfyJB48CvgkcXMJ6Fc3OxEyx9txz\nKXfe+SMOOuhgLrnkIvbcc2mkFk9yHFvnxURLmISxTVUfzHv9SxHZWKoKTZRNoWCK9fnPX0ZHx07q\n6upZvfp+duzYxtlnn1fpag1yBxd3MiYawiSM34vI3wG/wB+38U7gORHZG0BVXy1h/UKzA8sUy3Ec\nZs1qYNu2VlKpFG9/+1+yxx7zKl2tQa7rkLETIRMhYRLGWcHjp4dtPw2/T2PvSa3RBNnthyasNWv+\nwA03rGL+/IV89KMf5/LLv8Dee+/Dyy+/yLnnfoz3ve/kSlcRANexlrOJljEThojMUdVtqjp8idZI\nsgPLhHXTTd/i05/+HFu2bObSSz/LDTfcyIIFC+nu7uKSSy6KTsJwHTsRMpFSqIXxI/zLT7tFRK7D\nn1okC1yiqmuC7QuB2/FbKQ5+S+VSVf3BWJ8pxJruJqxkMjk46eAdd/yQBQsWAlBTU0sqlQr1HeWI\na9e122pNtBRKGM7ufrmIHA3sq6pHiMhy/CnRjwBQ1U3AO4L9YsBDwD2FPlOIHVhmImpra4e8dpzx\nw75cce061odhoqVQwthLRK4c601V/VKI7z8OuDvY/wURaRSROlXtHLbfecCPVbVbRMJ+Zgi7JGXC\n2rRpIzfd9K0Rzz3PY/PmTWG+oixx7boO2YzFtYmOQrPVekCmwL8w5gMtea9bg23DXYA/hXoxnxnC\n8oUJ673v/Stc18V13SHPY7EYJ54YalKDssR1zPowTMQUamGsVdUrJrm8Ee19ETkceL7AmVaoS2N1\n9VU0N9fvTt0mpBJlVrrsqf4zX3rppM9IW5K4dl0Hj5n3e57q8TWVyi1WSfswgE0MPYtayK51NXJO\nAlYX+ZkRdrR109JS3lG6zc31ZS+z0mXP1J95mLLEtes4ZDLZGfV7nqnxVcmfuRiFLkl9YPeqAsD9\nwAcBRGQlsFFVu4btcyjwZJGfGcGza1KmfMoS1/4lqUmrszG7bcyEoao7dvfLgwkKHxeRR/HnpLpI\nRM4VkVPydpsPbC30mTBl2d0kplzKFdeu65DNemRsBkITEWFGeu8WVb1s2Kanh71/UIjPjMs6B81k\n+N73buHss88fd79yxHVuHEY26xELs5iyMSU2bcLQWhhmMuy///JKV2FQbqS3xbaJigklDBH5wmRX\nZHdl7H51U4TsKJd50uk0hx329grUZnRx1yWT9WxNDBMZE21hPDGptZgEdp3XhLV58yY+8pHT6Ozc\ndcfrc889w8c+9hHa2toqWLOhYjF/enOLbRMVBROGiIx4X0QSqvqL0lVpYqzZbsL6+tev4/zzP05d\nXd3gthUr3sw553yU66+/roI1GyoedFz0py1hmGgYM2GIyDL8dS8a8ra9Df9Oj7llqFtR7JKUCWv7\n9m28+90njNh+3HHvYcuWcYf8lE0s5g+F6h8IO7GCMaVVqIXxNeAKVW3PbVDVPwBX4a/xHSlpa7ab\nkDKZsf8A9/b2lrEmheVaGH2WMExEFEoY81X1+8M3quoPgWUlq9EEWQvDhFVXV8dzzz0zYvuf/vQ4\nDQ2NFajR6GJBwhiwS1ImIgqNwyj0Xs1kV2R3WR+GCesTn7iQL37xUo4//r0sX76CbDbDU089ycMP\nP8A3vvHtSldvUHzwkpQlDBMNhVoYbUGfxRAicgz+TJuRYneSmLBWrHgzN954G67r8otf/IwHHvgl\ns2bN4tZb/4dFixZXunqD4m6uhZGx9V5MJBRqRfwj8GMR+R7wRyAGHIk/H84xZahbUeySlCnGnDlz\n+fjH/7bS1Sgo18LIeB7pjEciPhnzgRozcWMmDFX9g4gcij/nzdn4a2A8DRykqtvKVL/Q0pYwTEin\nnXYy+ZMxO47fr/Hud5/ImWd+pHIVGyYe91sYmYzHQDpLIj5tJmYwU1TBuaRUdQtweZnqslvSGbsk\nZcJZteqbI7Zt29bKXXfdwa233sR5511QgVqNlAg6vTNZj4F0hjJM/WZMQWNGoIisxV91L8cD2oHb\nVfXaUlesWP1pu/XQhDNaP8WiRYtZseLNXHzxJ6OTMBIxwD8ZssF7JgoKnbK8a5Rt84ELReSLqvov\nJarThNith2Z3xeNxEolEpasxKBm3hGGipVAfxiujbH5FRH4PPAxEKmHYAWV214YNr+E40eknyPVZ\npDNZ0uksnufhONbxbSqn6IuiqpoWkf5SVGZ3WAvDhPXlL1/O8BWIOzp2snbtWq688urKVGoUyUQu\nYXh4ePQNZKhKWj+GqZyio09E9gMi99d5wPowTEhvfethI7bV1NRw8MGHMGtWwyifqIxkXh8GQG+/\nJQxTWYU6vb/H0E5vgNnACuBDpaxUseIx11oYJrQTTzxp1O1btmzhzjt/FJlO72R8VwsDoKcvTWNd\nqpJVMjNcodOV1aNs6wAeVtXtJarPhCTiljDMxPT39/Pwww/ys5/dw0svKccf/95KV2nQ8BZG/0CW\nrOfhWj+GqZBCnd63jbZdRPYUkYtU9culq1ZxkglLGKY4zz77DPfe+xMeeugB9t57H954Ywt33nkv\nqVRVpas2aPC22iC2PTx6+9LUVEXnTi4zs4S6ICoiKeBU4HzgYOB7paxUsaqScXr70pWuhpkiPvKR\n0+jr6+c97zmBG2+8jcWLl3D++WdFKlkAVCf9hJF/B2BHz4AlDFMxBROGiBwGfBQ4DXgG2BNYoqo9\nZahbaFWpOO2dfZWuhpkiFixYyLPPPkNLy1a2bWtl8eIlkbxdNdfBnb+AUk9fmnQmO7hWhjHlVKjT\n+zmgCrgdeJuqviwif4pasgD/TKxvIGP3qZtQvvrV/6C1tYWf/eynXH31FTiOS29vD9u3b2P27DmV\nrt6gmqpcwhh6ubWje4Cmeuv8NuVX6DRlLdAALAYWBNsiOcNfdSqO59ngPRPe3LnNnHPOR/nhD+/m\nH/7hMg455K2ceeYHuPzy/1fpqg2qTvmXnvqG3TLe0d1P1tZ/MRUwZsJQ1fcBBwIvAbeIyIvAPBGZ\nV67KhVWV8s/E+vptLIYp3sqVb+Xyy7/MHXf8HytXvhWA++77vwrXCqqrRl6SAsh6Hm12CdZUQMEL\noaq6SVWvVtV9gU8CDwIvisj/lqV2IdUECaO33zq+zcTV19fz13/9QSAiCSPX6T3Kinsd3QMjEokx\npRa650xVH1LVs4GlwEMAInJOqSpWjKrBhGEHkJkcUVjhLhZzicecURODh8e2nb0VqJWZyYq+1UJV\n21T1P4OX501udSamttq/1ttjt9aaSRKVmydSCf+GjtH0DWTYbknDlNHuTkwz7lElItcBh+PPP3WJ\nqq7Je28x8H0gATyhqheKSC3wXaAJSAJXqur9hcqor0kCfjPdmHIoR1yDf0NHV+/YJ0I7u/uJx1xm\n1SZ36+cxJozdvZm7YLtdRI4G9lXVI4ALgK8P2+Va4KuqejiQCQ6084AXVPWd+OM//mO8SuQOlvau\nyE2ia6ahcsU1+LfW9vanC14i297RS1evnSyZ0iv16J/jgLsBVPUFoFFE6gBExAGOBH4avH+xqr4O\ntAK5m+FnAy3jFZJLGB3dljDM5KitrS30dlniGnbdMj5e/1xrWy877YTJlFipE8Z8hh4YrcE2gGag\nE1glIo+IyNUAqvpDYKmIvIS/UNPfj1fIYMLosbMsM75sNsutt95EJrPrj/D69eu47bbvDL7+ylcK\nrkJclrgGqA2mAekucFkK/E7w7R29bN3RTSZr45FMaexuH8bOIvd3hj1fBHwNeA24V0ROxD/7Wq+q\nJ4rIW4DvAIcW+tJcH8ZAxqO5ub7IKu2ecpcXhbKn+s98/fXXs379KzQ0pKiurgagunovvve9tdx3\n312cc07RN/+VJK4BZjf49Ysn4jQ1Fmz1DOrNOsxrrKY6tXuHt8XX9C+3WIWmBnGBy4CvqGom2LYc\nOFVVrwJQ1feP8/2b2HXmBbAQ2Bw8bwXWqeq64LsfAN4M7AX8Ivj+p0RkoYg4qjrmRdz6oIXRuqOb\nlpaOcao0eZqb68taXhTKng4/8y9/+QD/+Z/fobMzTWfnru/7/Oe/yCWXXMiJJ/71iHKHKUtcA7hB\nN+GWbZ001IRPAC2tHVSn4jTUJie06JLF1/QvN1d2MQpdkvoSsBLIn7RmE3CQiHw65PffD3wQQERW\nAhtVtQsgSEKvisg+wb6HAC/gjyw/PPjMUqBjvIOqtipOPOaw0/owTAipVIpkcuRdRalUVdg1vcsS\n1wB11f4f+64JXG7t6UuzZXs3m7d12S3nZlIUOjpOAs5U1e7cBlXdCZxLyBX3VPUx4HEReRRYBVwk\nIueKyCnBLn8H3CoivwHaVPWnwLeBZSLyMPDf+CPMC3Ich8a6FO2d/WQjMODKRFtPTzc9PSPn0Ny5\ncyfd3d2jfGKocsU1QEOwwt7udGj3DWR4Y0c3m1q7aO/ssxHiZsIKtVV7VHXEhDWq2iMioXvVVPWy\nYZueznvvFeCoYft3MYElYBvrUrS299Ldk6auxtYLMGM7/vj3ctllf89nP3spS5bsCcBLL73Iddf9\nG2ec8eFQ31GuuG6o9RPGZNwy3p/O0N+ZYUdnH3HXpToVD/7FIjNQ0URboYRRJyK1uaZ2jog0AZHr\noclN99zS3mMJwxR0xhkfIZFI8pnP/C2dnZ14Xpamptmcc875nHDC+ypdvSEa6/xLZ5N9y2w6m6Wj\np5+Onn4cHKpSMZLxGKmESyIem9SyzPRRKGF8D7grWI71JQAROQi4AX9gUqTkEsbWHd3stWBWhWtj\nou7UU0/n1FNPp6urE8dxqampqXSVRpWIu9Sk4iUdY+Hh0dOXHtLP0ZuFzo4ekvEYyYRLMh4jkXBt\nPfEZrtCa3teJSB/woIg04Pd3vAFcraqRWqIVYPasoIXRZnPrmPG99to6brvtZl588QVc1+VNbzqA\n8867gPnzF4z/4TJyHIdZtUla23vLukBYNuvRN5Dx57HK6+6Juf6EiLGYS9x1iMfyXsccYq6tBDid\nFbzfTlVvAG4QkVlAVlU7y1Ot4s1r8s8Qt7b1MJDOkohb4JrRrVnzB77ylSs588yPcNppZwDw3HPP\ncsklF/GP//hPHHjgQRWu4S4Ofut5y/ZuOroHKj5nVCabJZMFxug4d3CIx/xEEgsSiOs6xFwH1wke\n3V2PZmoZb01vAf4R//barIj8Afiyqq4vR+WKsbjZH9TU0tZDb3+aRNwmYzOju/XWm/jqV1ex9977\nDm5bvnwFhxxyKNdd92/8x3/8Z4FPl5fjQHNjFc+v90+GKp0wxuPhMZDxGMiMf1+Mg580XIfBJOLF\nYrR19A2+dhxwHf/RcZzB565jCacSCg3cOw64Gfh3dk2u9jbglyJynqr+tgz1C62pPkV1Kk7Ljh56\n+tKDo7+NGa6vr29IsshZunTZqLfbVpTjsEeTP9q7pa2HfRc1VLhCk8fDI5P1yAAEDZadXf20d4Vf\nTdAhSCCuEySU0ROMEySYXLIB8PNN7n3o7UvTP5AZ3B8YfO4QnSnvK6lQC+Ny4H2q+kzetjUi8iB+\nx/dxJa1ZkVzXZV5TNeu2dLCza4DZs7LEY3ZZyoyUTo89CK7Qe5XgAns0BgljR8SSWQR4eHgeZDO7\nP/6qL+uwo62r4D4OzpAkwpCEMvQ5juMPdHN2fW7we/ISUiyVoL2zD/K+M/eQv9+ut4Z/l//9wf/y\n9i+830R8lrSQAAAaCUlEQVQSYKGEUT0sWQD+7JzB3P6RM292Deu2dNDS3k1zY9XgoCdj8i1cuIhH\nHnmYo446dsj2Bx64nyVLllamUmNwHIfZs6qIuQ5b2yxhVFouQTGJA4SdeJwdFVijva4qUfTUIIUS\nRqFrOpG83rNort/xvbGliyXN9ZYwzKguvPAzfO5zF7N69f2sWHEA2azHM888ybp1a7nhhpsqXb0h\ncpdbmhur2bqjh3Q6S9xu6DAVUijyXs2b6mCQiHwI0NJVaeL2X9IIwNrNHaSzWbptURkzikWLFvPd\n7/6QQw45lA0bNrB9+zaOOupYbrvtBzQ2Nla6ekPkLhssnV9HOuOxoSWyNyqaGaBQC+MfgJ+LyBnA\nH/CTyxHACoZNexAV85pqmFWbZN2WDjzPY3tHH1WpuA02MiMkk0mOOeYdnHzy0JlpN2/exIIFCytU\nq7HttWAWv39uK2s377SBqaZixmxhBPPhvBl4ENgPfzrnu4G3qGpreapXnGQixl4L6oNZOntIZ7K0\nd9oMtmaoJ5/8E+9//4mceeapnHXWqWzc+DoAP/7xD7nwwgsqXLuhcneOLp1fj+s4rN1UmWmwjYHx\nB+71icidqnpj/nYRWZab7z9KEnGXvRbM4smXt/HqpnYWzKlhZ1c/ddUJG8hnBn37299k1apvsmzZ\nXvzmN7/immuuIpvNUl8/ixtvvK3S1Rsid0kqlYixsLmGjS3+VOW7uziSMRMx5l9RETlKRDYCL4rI\nC7n5/UXkU8BvylXBYsRjLvstbiDmOjz58jY8z8PDo6Wth2zWpj03Ptd1WbZsLwCOPPIYNm/exAc/\neAZXX/1V5s5trnDthsq/mrr/4kY8D55Zu71yFTIzWqHT7quAd6nqHPz+jG+LyEPAO/EH8EVSY12K\n5Xs20trey4atfgdhfzrD1rYePFsrwzDy/vN58+ZzzDHvqFBtCsuv61/sOxfXcXhCWyyWTUUUShgZ\nVX0eQFXvAZYBX1fVD6jqpnJUbiKS8RgrxT9LfOLFXV0tvf1pWtptYkIzUpRH8ObfsFFXk2D/PRt4\nY0cPm1oLDzAzphQKXQgdfgrzmqreVcrKTIZUMsay+fU01ad4bt123vXWxdRV++tjdPcO8MYOj7kN\nVTar5gz2zDNP8YEP7Fr3oq1tx5DXd955byWqNarhuWzl/s28sL6NP77QwqLmuspUysxYxfScTYk2\ncFXSXz3s8APmcd/vXuOBx1/nlCP3Gny/py/N5tZu5jZWUZW0jsOZ6H/+58eVrkJowyfY22fhLJob\nq3jqlW0c+qY9WDQ3kpMumGmq0F/MI0TktbzXewSvHcBT1T1LW7WJicdcEjGXQ/Zv5glt4cmXt7Fy\n/2aW7LHrbCydzfLG9h4a6pI02mjwGSdqa14UMnxCVsdxOPGwPfnuL17kvsfW89H3vclmbTVlU+i6\njOAP0Mv9y70+kogO3MupTsVxXYcTD/dz2n2/Wz/iLikPj7bOPja2dNLR3W+diCaSRutfWbZgFgfu\nPZtN27p54sWWCtTKzFSFVtyL3JoXYVUl4+zs7mfPefW8ZZ85PPXKNn7++9c48fA9RxyAA5ks23b2\n0t7Zz6zaJHU1CRsZbiJjrFh811sX8+KGdn655nUWzKmx/gxTFtOy57c6FSMedGqfcNgS5jVVs0Zb\n+PWTm8f8TDqbZXtHL69v7WT7zl56+9PW6jAV57oODiOTRn1NkvcftYx0Jsv/rH6ZVrsD0JTBtEwY\njuNQG9wZVZWMc9a796OxLsmv/ryJx7VwEz7reezs7mfL9m42bO2kpa2Hrt4BG/hnKmasBq/s2cT7\n3r6Unr40t9//Iju7bBocU1rTMmEA1Nck8p4n+fB79qemKs69j61n9ZrXyWTHX0Iy63l09Q7Q0tbD\nhq2dbNnezY6OPrp7B0iHWILSmMkQK9CpvXL/Zo49eCHtXf3c8rMXBgerGlMK0zZhxGMuNXnz7cyZ\nVcXZ79mfpvoUv31mC7fdp7QVsWiJh0dvf5r2rj62tvXweksnaze1s2V7N9t39tLZM0D/QMYuY5lJ\nN95dUEe9ZQHHHryQnd393HrfC/zmqc0Wh6YkpvVAhKb6Knr6uvCCISTzZtfwib9awf89tp5n127n\n2/c8x/vevpQVy5omNNo3k/GTSG/elQAHh3jcJR5ziLv+Yyy41TcWc2zZWFO0Qi0M8C/BHn3QQvbc\no467HlnLg09sZN2WDv7qL5fRUBvJtc7MFDWtE0Yi7jKrNjlkUflUMsYHjt6LvRbU8/Pfb+DHv3qV\nR56q5sgDF7BiWdNu39Pu4TGQzjCQHv19B2cwccRdP7m4rkPM9Resj7nO4OsoT1lhyidsTC5bMItP\nnLyCn/xmHS+/3s71P36ag/aZw18eOJ/Zs6pKXEszE0zrhAHQUJekq2eAdF6fheM4rNy/maXz6nnk\nqc08/eo27vz1qzz85xRHHriAA/eZXbKpQzw80hkvVB+I6/jJY3gicV2HZHWSrt6BvH38n8t1di1S\nb6aHYmKxtirBmcfty1OvbOM3T23hTy+18ueXWzlgr9kceeAC9miqLmFNzXRX8oQhItcBhwNZ4BJV\nXZP33mLg+0ACeEJVLwy2fxj4PDAAfElV75to+a7jr4e8ZXv34KWpnDkNVbz/qL045i8W8ujTm/nz\ny9u459F1PPD46+y/pJH992xk7wX1JOKxiRa/W7KeRzYz+rVoJx5nR1vPmJ918BNHLoG4QYtlyDbH\nCZIMOK4zbPuu78hPQJlMlqzn4TCzk1I543q8S1LDOY7DQfvO5cC95/D8+h385qnNPPPqdp55dTtL\n9qhj/yUN7L+kkbkNVTP6d2iKV9KEISJHA/uq6hEishy4GX+Z15xrga+q6j0icn1woHUDXwIOBuqB\nK4AJJwzwL0Pt0VTN1h09I5IGQFN9ipOOWMbRBy3ksWe28PTa7fzppVb+9FIr8ZjL3gtnIUsa2G9J\n4+BEhlHn4eF5ftIBIDM539s14LGjbddMqbmkAsEgsyC5+All6HPytzO0JeTmvR988YikVN3dH6zT\n7m/LvZX7vqGfI+R+xf/BLHdcx2MT+6Puug4H7DWbFcuaeOn1dh579g1ee6ODDVs7eeDxjTTVpwaT\nx57z6mxCTjMup5R3U4jIFcB6Vb05eP0c8DZV7RQRB3gdWKyqXt5nTgeOVtVPFVGU19Iy/tKVnT0D\ntLaPfVaek816bGzt4sUNbby4oY2Wtl2DovZorGb+nBrmz65mvz3nUJt0qKrA6mdNjbVD/nBP93JL\nXbYzLLn4z/2tBx+wYMhf7HLHdd9Ahs3bJufn7u4d4KXX23lxQzuvbGqnf8C/NJpKxFgwpyaI7Rr2\nXzqHpOuVfZ6q6RpfUSy3rirBm/bbo6hfcKn/0s0H1uS9bg22vQw0A53AKhFZCfxaVf8Rf92NWhH5\nCdAIXKGqD05GZeqqEzgObGvv3XXmPQrXdViyRx1L9qjjuEMWs31nb5A82tnY2sXWth6eegXu/6O/\nFnRjXXLwQJs/u4bmxmpm1SbsjG0KybU8h4TF2DFS1rgu9pJUITVVCQ7ady4H7TuXdCbL+i0dvLih\nnVc3tbNuSwfrtgQnXo+sJeY6zGuqHoztebNraKpPUVsVt0tZM1S5T42dYc8XAV8DXgP+T0TeG2yf\nDbwf2At4CFg6WRWorUqQjMdoaeuhPx3uOs3sWVUcfsB8Dj9gPtmsx/aOPrZs66ata4B1m9vZsq2b\nF9a38cL6tiGfq69J0FiXoqEuSUOtPzNuQ20ymCU3WbG+ETPpShrX8ZiLgzPq5dTdEY+57LOogX0W\nNQDQN5Dhje3dbNnew47Ofl7bstNfrGlb97DPOUEcp2gMYrshiPPG2iT1NUmbQXeaKnXC2IR/5pWz\nEMhN6NQKrFPVdQAi8iCwAtgC/DZozr8qIh0iMldVWymgubm+qIotmD+L1rZe2osYvJczZ3Yd+y2d\nM/ja8zzaO/t5vaWDjVs7aW3vZfvOXnbs7GVjSycbto7+PbXVCZrqUzTUpaitSlBbHQ8eE7seg23V\nVYkRZ5pNjZVZC6FS5Va67Dxlj+vutMdAuvSzC8xvnsVBea/TmSxvbO9mY0snm1u72L5zV2xv27Rz\n1O9wHWioSzF7VhV1NQlqqhLUVfuPfmzHh8R4bg2bnJkYX5Uot34CY3RKnTDuB/4ZuDFonm9U1S4A\nVc2IyKsiso+qvgIcAvwP8Dhwq4hcg39GVjveQQUQpg9jNFUxaO/KdaYWZ/i1x0Wzq1k0e+hti9ms\nPzdVe2c/7V39tHX2DXm+ubWL10NO51CVjFGTilNdFaehLkXCdaiuilOVjJGMuyQT/mMieMzflnuM\n7ebAwZl4jZkljcO3lD2uOzp66OkbY3BPCeT/t65JOOy3sJ79Fg49KesfyARx3E97Zx/tXX6ct3X5\nMf7KxvZQZbmOQ3UqRk1VnPraFMmYQ3UqTnUqTmp4bOfH9JCYd3f7EvBM68MY6O1n3uyaoj5T0oSh\nqo+JyOMi8ij+fToXici5QJuq/gT4O/yDyAGeVtWfAojIj4Df4a/yV0wnYdFSiRh7NFbTP5AMEkd6\nUpv+ruvQWJcac6Emz/Po68/Q3Zf2//Wm6Qme9wy+ztDdO+A/9qVp29bNxpaJBZjrOCQTuw62xCiJ\nJREcmPGYS8x1Bkerx12HWfWd9PX1E3ODketu8F7MIZY3sj0ejBnJvTedrnlXIq6T8VhZE0YYyUSM\n5sZqmhtHH9uRyWbp6cvkxXGhGE/7N6W09U746Iu5udgO4jp4nky4JOK5+Pbjenhsx3Kx3ds/GMu5\nAbZ+HPszNwx5DPabSZffSnqXVBmFuksqjIF0lo7ufrr70uMOrqvUmYHnedTUVLFp6066ewfoG8jS\nP5ChP+0/DqSzg8/DbCvHpQ7XcYYcZEMTzK4DNH+wohuMH8ltq0olSKfTg68dx8F1hw5wzA1iHPq6\n0PbxP//eo/at1F+EwbgOe4ffZKlUbDfMqmHz1na6g0TTl4vTArE8MJClP52hP+/R3ydDeoxxTJPJ\ncciL6SDJ5CWWEXE9LPb8uM6MiLuC8R0ytnODeUfbp6E2ybFvWxapu6SmnETcZfasKmbjN7lzZ0J9\nA5M0kGESOI5/K29TfYqm+t1fYjab9RjIjDwA0xmPTDZLJuMNPk+lkuzs7CWTyZLOev5j/n7DtqUz\n3uC+2aw/wj2330B/esh+UfXeo/atdBVIxGfGHXeu61BT5fd3TIZs1qM/nUs6WfoGMn4MZr3B2MwE\ncZlKJeno7CU9GPN5+wXbMkNieOhjJi/2+weydGfSg/tH9bz82LctK2p/SxgFJBMxkokYDXUp0pks\nPX1pevv9P6rpMpyVl4vrOqTcGKnE+HdtlerM0/N2DTTMZr3gcejr+rpq2tq7h+3jkfUY9nrkZ8fe\nb+zPe57/RyAKEvHS3Ck13bmuQ1UyTlWI/t1StqoKxXd9fTU72rrIZkfbp3B8Fh/bu15XhTjeh7OE\nEVI85lJfk6Q+6CPKeh4NDTU42Yx/Zh40g+2AnpjBaUtwYIw4bmqsJsb0SdTFcB2HRNwNfSu4iZZC\n8d1Yl8JLl79/qm4CrThLGBPkBpeFZtXsOnXxPP/Wx4HMrksx+c8LDRY0ZjzVqbglDFNRljAmkeM4\ng5exRpPrK8hd50xnsoPXSTPZXU1JY0ZTlYzRXpk7mo0BLGGUVa6vgALXDvOvS+aSyJDH4L2qVIxE\nzCUTXPe0S2HTXyoZs34MU1GWMCLGcRxijkPM9efGHktzcz2pvBviconGCzq1cp1n/naGPO7aJ+8z\no7wPloiixHUcUskYvf3RGo9hZg5LGNNELtEAY3YaT1QuiXh4zJlTT03cAfwEg7drKnUv7zkEU6v7\n/9v1Hd6uSf48z09Huffz991Vtr8NzyMRDLoaaz+Y/gmurjphCcNUjCUMM67cHR7gDI4Cr4Tm5npq\nQq4NkUtG5CWSwaQymFO8vIRUaL/oqKmK4+50rK/LVIQlDDMt5dayGLIi0zTgOg61VQk6evorXRUz\nA82M4aPGTCP1NVNj1Ucz/VjCMGaKSSZiU2apYDO9WMIwZgpqqk/5a6gbU0aWMIyZgmKuOykTTxpT\nDEsYxkxR9TXJSZvV1ZgwLGEYM4U1N1RRnbSbHU15WMIwZgpzHIfmpupQU9Mbs7ssYRgzxbmOwx5N\n1dbSMCVnCcOYaSDmusybXUNTXQpnmgxSNNFjCcOYaaShLsX82TXEXTu0zeSzqDJmmkklYyxsrqWp\nzsZqmMllFz2NmYZcx6GhLkV9bZLO7gHau/rJZGfm8rZm8ljCMGYacx2HWbVJ6msSdPQM0NUzQN+A\nLfNqJsYShjEzgOM4zKpJMqsmSSabpacvQ3fvAD19mWm/hoiZPJYwjJlhYq5LXbVLXXWCrOfR25em\nuy9N/0CWgXTWEogZkyUMY2Yw13GoqUoMTjHieR796Sz1s6pJ9w3Qn87QP2BJxPgsYRhjBjmOQyoR\no6EuRX9DFeAnkXQmS99AlnQm98/flsl4lkxmEEsYxpiCHMchEY+RiI8+/cjwJJJLJAOZLNmsZ8vJ\nTiMlTxgich1wOJAFLlHVNXnvLQa+DySAJ1T1wrz3qoBngCtV9bulrqcxxbC43iUec4nHxh7S5Xl+\n0shmPTLZMR6D9xMJF9exNcujqqQJQ0SOBvZV1SNEZDlwM3BE3i7XAl9V1XtE5HoRWayqrwfvXQ5s\nK2X9jJkIi+viOI5DzHGIuX4GLaS5uZ6amD/YMJPN4nmQzXr+o+cFyYehj8H7ucSU2zf3/uB32KWz\n3VbqFsZxwN0AqvqCiDSKSJ2qdoqIAxwJnBG8f3HuQyIiwHLg3hLXz5iJsLgug1huepNJnIg3l3T8\nBAOw6/mcuXUkHW9wezbIL4P7+y9G2e6/mQ3ez+07/HOD6SrvfQDXdYbM/xXlxFbqhDEfWJP3ujXY\n9jLQDHQCq0RkJfCIql4W7HctcBFwXonrZ8xEWFxPUa7jwBjTpVSn4lSnyt+t29xcT31y6CU9b0SC\nySWy3PvBG8HzvJdj7usNPvfFY8VPG1Pu/zrOsOeLgK8BrwH3isiJwFzgt6q63j8hs6k3TeRZXJtJ\n5Th5bQ5n8P8qrtQJYxP+mVfOQmBz8LwVWKeq6wBE5AHgzcBKYG8R+StgMdArIhtU9cEC5TjNzfWT\nXfdQKlVuJcueiT/zMNM+rsHiayaUW6xSz1Z7P/BBgKB5vlFVuwBUNQO8KiL7BPseArygqmeq6mGq\n+nbgJuDL4xxUxpSbxbWZkUrawlDVx0TkcRF5FMgAF4nIuUCbqv4E+Dvg1qCj8GlV/Wkp62PMZLC4\nNjOV43nR7ZE3xhgTHbaAkjHGmFAsYRhjjAnFEoYxxphQpvTkg4Xm8ylD2dfgj+iNAf+qqneVseyK\nzEckIh8GPg8MAF9S1fvKUGYt8F2gCUji/8z3l7jMN+OP5L5OVb8ZzA31PfwTrM3A2ao6UOI6WGyX\nKbYrEddBuVMutqdsCyN/Ph/gAuDrZSz7WGBFUPaJwKpylR0o+3xEIjIb+BL+nEknAaeUqejz8G9L\nfSdwGvAfpSxMRGrwY2l13uYrgetV9RjgFeCjJa6DxXaZVDCuYQrG9pRNGAybzwdoFJG6MpX9K/xf\nMEAbUBPcQllyFZyP6F3AL1W1W1XfUNW/KVO5rcCc4PlsoKXE5fXi/6HcnLftWCB3a+xP8f9blJLF\ndvlUKq5hCsb2VE4Y8xn6Hzg3n0/Jqaqnqj3BywuAn6lque5Pvhb4LOWfK2AZUCsiPxGRX4nIO8tR\nqKr+EFgqIi8BDwN/X+LysqraN2xzbV4zfSuwoJR1wGK7nLG9jArENUzN2J7KCWO4sk+2IiKnAOcD\nnypTeWcTzEcUbCrnz+zgnwW9H/9nvqUchQbXl9er6n74Z943lKPcAioxqY/FdulUJK5hasb2VE4Y\nhebzKTkROR74AnCCqnaUqdj3AaeIyGP4Z39fLOMZ0Rv4B7Snqq8CHSIytwzl/iXwCwBVfQpYWK5L\nJHk6RCQVPF+EH3ulZLFdvtiuVFzDFIztqZwwxpzPp9REZBZwDXCSqraXo0wAVT2jgvMR3Q+8U0Qc\nEZmD35RtLUO5L+PfLYSILAU6yniJJGc1cGrw/FTg5yUuz2K7fLFdqbiGKRjbU/a22tHm8ylj8R/C\n76z63+CMwAPOyVtVbdpR1U0icgfwO/yftyyXKoD/Am4WkYfxb/P8ZCkLC/5AXwssBQZE5IPAh4Hb\nROSTwHrgtlLWwWK7fCoY1zAFY9vmkjLGGBPKVL4kZYwxpowsYRhjjAnFEoYxxphQLGEYY4wJxRKG\nMcaYUCxhGGOMCcUSxjQkIktFZMM4+/yTiFxZrjoZMxkstivLEsb0ZQNszHRlsV0hU3akt/EFo3G/\nBQiQAn4PfC3v/VuAHmBv/PmJblXV3BoHS0TkR/hTSj+sqhcHc+bnFnWpB+5Q1WvK9fMYk2OxHT3W\nwpj6moAnVfXYYB6e44HhaycsVNUTgGOAy0WkKdi+D3A68Fbg3GD7HsBdqnoc/qprl5VxLQZj8lls\nR4y1MKa+NmBPEfkt0Id/pvXWYfvcD6Cq7SKiwH7B9t8Ek531icg2oBF/9s6jReRCoB//zG420Fny\nn8SYoSy2I8ZaGFPfGfgH0V+q6jvwZ8Aczh32PHcNOJ233cOfD/8SIKmque+zg8lUisV2xFjCmPrm\nAaqqnogcgt8UTw3b5x0AQbN8H0BH+Z7cPPzzgOeC/U8Gqkf5PmPKwWI7YixhTH0/Ao4QkYeAvwb+\nHX+h98a8fXaIyF3AQ8CXVHXnKN+TOzO7GThfRFbjT4N8e/DPmHKz2I4Ym958mgvuJHlEVW+udF2M\nmUwW2+VnLYzpz84IzHRlsV1m1sIwxhgTirUwjDHGhGIJwxhjTCiWMIwxxoRiCcMYY0woljCMMcaE\nYgnDGGNMKP8fhaBTb0gFNBsAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb68eefa110>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_with_err(ax, x, data, xlabel, ylabel):\n",
    "    mu, std = data.mean(1), data.std(1)\n",
    "    lines = ax.plot(x, mu, '-')\n",
    "    ax.fill_between(x, mu - std, mu + std, edgecolor='none',\n",
    "    facecolor=lines[0].get_color(), alpha=0.2)\n",
    "    ax.set_xlabel(xlabel)\n",
    "    ax.set_ylabel(ylabel)\n",
    "\n",
    "@use_result('my_load_data')\n",
    "def task_validation_curve_alpha_rocauc(train_features, train_targets, train_site_cnts, **kwargs):\n",
    "    alphas = np.logspace(-2, 1, 25)\n",
    "    \n",
    "    sgd_logit = SGDClassifier(loss='log', random_state=17, n_jobs=-1)\n",
    "    \n",
    "    trunc_train_features = train_features[['day_of_week', 'start_6_12']]\n",
    "    X = sparse_hstack( (csr_matrix(trunc_train_features.values), train_site_cnts) )\n",
    "    y = train_targets\n",
    "    \n",
    "    val_train, val_valid = validation_curve(sgd_logit, X, y, 'alpha', alphas, cv=5, scoring='roc_auc')\n",
    "    \n",
    "    f, (ax1, ax2) = plt.subplots(1, 2)\n",
    "    \n",
    "    plot_with_err(ax1, alphas, val_train, xlabel=r'alpha', ylabel=u'ROC_AUC. Тренировка')\n",
    "    plot_with_err(ax2, alphas, val_valid, xlabel=r'alpha', ylabel=u'ROC_AUC. Валидация')\n",
    "    plt.legend();\n",
    "    \n",
    "task.call('validation_curve_alpha_rocauc')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Вывод.** Я не исследовал глубоко требуемое значение *alpha*, видно, что главное — лучшее $\\alpha$ в районе долей единицы, причём практически разницы в параметре ROC_AUC при *alpha* < 0.1 нет (**ROC_AUC** в районе 0.74). Поэтому примем данный параметр по умолчанию ($\\alpha = 0.0001$).\n",
    "\n",
    "Тем более ответов на данных, которых мы будем предсказывать, нет, поэтому не будем сильно замочариваться глубокой настройкой параметра, который слабо влияет на качество модели в определённом интервале.\n",
    "\n",
    "Выбранное значение $\\alpha$ можно подтвердить или опровергнуть на кривых обучения (следующий подраздел). Если модель при выбранном значении $\\alpha$ продолжает обучаться при увеличении количества данных, значит, $\\alpha$ приемлимый."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 10.3. Кривые обучения"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Кривые обучения показывают **поведение метрики качества обучения при увеличении количества новых данных.** По оси *y* — метрика **ROC_AUC,** по оси *x* — размер обучающей выборки. В идеале при увеличении размера обучающей выборки кривая на тесте должна приближаться к кривой на валидации, при этом кривая на валидации должна подниматься всё выше и выше (качество на данных, которые модель не видела, должно расти при увеличении размера обучающей выборки до некоторого порога, когда модель уже перестаёт обучаться). \n",
    "\n",
    "Анализ кривых валидации позволяет не только установить корректность модели, но и оценить потенциал модели обучаться дальше (если будут получены ещё данные). Если кривая валидации пришла к некоторому верхнему пределу — новые данные вряд ли улучшат модель.\n",
    "\n",
    "Построим кривые вадилации по двум метрикам: по **accuracy** и по **ROC_AUC.**\n",
    "\n",
    "**Примечание.** Посмотрим, как сближаются кривые при разных $\\alpha$."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAuYAAAHyCAYAAACnNgtDAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nhNlJebI64wAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb654308690>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_with_err2(x, data, label, edgecolor, facecolor):    \n",
    "    mu, std = data.mean(1), data.std(1)\n",
    "    lines = plt.plot(x, mu, '-', label=label)\n",
    "    plt.fill_between(x, mu - std, mu + std, edgecolor=edgecolor,\n",
    "    facecolor=facecolor, alpha=0.2)    \n",
    "    \n",
    "def visualize_learning_curve(alpha, train_features, train_targets, train_site_cnts):\n",
    "    sgd_logit = SGDClassifier(loss='log', alpha=alpha, random_state=17, n_jobs=-1)\n",
    "    train_sizes = np.linspace(0.05, 1, 20)\n",
    "    \n",
    "    trunc_train_features = train_features[['day_of_week', 'start_6_12']]\n",
    "    X = sparse_hstack( (csr_matrix(trunc_train_features.values), train_site_cnts) )\n",
    "    y = train_targets\n",
    "    \n",
    "    N_train, val_train, val_test = learning_curve(sgd_logit, X, y, train_sizes=train_sizes, cv=5, scoring='roc_auc')\n",
    "    plt.figure(figsize=(12,8))\n",
    "    plot_with_err2(N_train, val_train, label=u'ROC_AUC. Тренировка', edgecolor='lightgreen', facecolor='green')\n",
    "    plot_with_err2(N_train, val_test, label=u'ROC_AUC. Валидация', edgecolor='lightblue', facecolor='blue')\n",
    "    \n",
    "    plt.xlabel(u'Размер обучающей выборки (alpha={})'.format(alpha))\n",
    "    plt.ylabel('ROC_AUC')\n",
    "    plt.legend()\n",
    "    \n",
    "@use_result('my_load_data')\n",
    "def task_learning_curve_alpha_rocauc1(train_features, train_targets, train_site_cnts, **kwargs):        \n",
    "    visualize_learning_curve(alpha=0.0001, train_features=train_features, train_targets=train_targets, \\\n",
    "                             train_site_cnts=train_site_cnts)\n",
    "    \n",
    "task.call('learning_curve_alpha_rocauc1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAuYAAAHyCAYAAACnNgtDAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nJDUBg7kkSZLUBP4vUapnyRX+p88AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb648028ed0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "@use_result('my_load_data')\n",
    "def task_learning_curve_alpha_rocauc2(train_features, train_targets, train_site_cnts, **kwargs):    \n",
    "    visualize_learning_curve(alpha=0.00001, train_features=train_features, train_targets=train_targets, \\\n",
    "                             train_site_cnts=train_site_cnts)\n",
    "    \n",
    "task.call('learning_curve_alpha_rocauc2')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAuYAAAHyCAYAAACnNgtDAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSZIkSZIkSdILyMA8SZIkSZIkSXoBbwMjrSuPrn4F1gAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fb6479bac10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "@use_result('my_load_data')\n",
    "def task_learning_curve_alpha_rocauc3(train_features, train_targets, train_site_cnts, **kwargs):    \n",
    "    visualize_learning_curve(alpha=0.001, train_features=train_features, train_targets=train_targets, \\\n",
    "                             train_site_cnts=train_site_cnts)\n",
    "    \n",
    "task.call('learning_curve_alpha_rocauc3')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Вывод на основе кривых обучения.** Видно, что более или менее приемлый параметр *alpha = 0.0001* или *alpha = 0.00001.* Причём параметр по умолчанию *alpha = 0.0001* мне нравится больше из-за того, что кривые обучения более монотонные, с меньшими возвратами к более низким значениям метрики, чем при *alpha = 0.00001.* Монотонность возрастания метрики на валидационных данных при увеличении количества данных — хороший признак (модель постепенно улучшается при поступлении новых данных).\n",
    "\n",
    "При *alpha=0.001*, во-первых, и сами значения метрики хуже (ось *y*), и модель очень быстро перестаёт накапливать информацию — уже после примерно 25000 строк кривая валидации указывает на отсутствие дальнейшего обучения — метрика **ROC_AUC** начинает случайно \"гулять\" вверх-вниз."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 11. Оценка модели по метрикам. Возможность применения модели"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Приблизительные метрики полученной модели:\n",
    "* **Accuracy:** 0.991\n",
    "* **Precision:** 0.694\n",
    "* **Recall:** 0.103\n",
    "* **Roc_auc:** 0.951\n",
    "\n",
    "В связи с тем, что модель имеет неплохую метрику **precision,** она не очень часто ошибается, давая ложноположительный результат (т.е. определяя пользователи как Элис, если он на самом деле не-Элис). Т.к. данные несбалансированные, сессий Элис не очень много, не удалось достичь хорошей метрики **recall.** Останутся сессии Элис, так и не найденные моделью.\n",
    "\n",
    "С другой стороны, если стоит задача обнаружить **не все** сессии Элис, но принадлежащие точно Элис, построенная модель окажет реальную помощь."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 12. Модель на практике. Что ещё можно сделать"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Данный классификатор можно легко приспособить для любых типов пользователей, в том числе для идентификации пользователей, которые тратят слишком много времени на нежелательных сайтах. Просмотр посещенных сайтов может осуществлять робот, что и надёжнее для работодателя, и этичнее по отношению к сотрудникам. Не отвлекаясь напрасно на списки посещенных сайтов, не тратя время, можно получать уведомление лишь в том случае, если Ваш сотрудник явно нарушает рабочий режим.\n",
    "\n",
    "Для этого нужно обучить модель на данных конкретного сотрудника. Это может потребовать дополнительной настройки параметров вручную. Но, выполнив их один раз для конкретного сотрудника, мы можем применять модель для дальнейшего его обнаружения, уже не изменяя её."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 13. Чего сделать не удалось"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Как было упомянуто выше, модель имеет сравнительно невысокую метрику полноты обнаружения всех сессий пользователя. При накоплении большего числа данных по пользователю можно будет увеличить данную метрику."
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}