notebooks/SALARY_01_TABNETL_PREPARE_DATA.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "SALARY_01_TABNETL_PREPARE_DATA",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hbUoifIyW1xO",
        "colab_type": "text"
      },
      "source": [
        "# Подготовка нашего датасета\n",
        "Данные были скачаны с использованием API hh.ru\n",
        "\n",
        "В условиях пользования серисом и данными hh.ru говорится:\n",
        "\n",
        "_1.5. Конечными пользователями Приложения могут быть юридические и физические лица, являющиеся как работодателями или  организациями, оказывающими услуги по поиску и подбору персонала и/или по поиску работы и трудоустройству (кадровые агентства), так и соискателями. В случае, если конечными пользователями Приложения являются организации и физические лица, являющиеся работодателями или кадровыми агентствами, то указанные лица вправе использовать Сервис только для целей привлечения потенциальных работников и их последующего трудоустройства непосредственно в такие юридические лица (у таких физических лиц) или у клиентов таких кадровых агентств._\n",
        "\n",
        "То есть мы вполне можем использовать эти данные в работе.\n",
        "\n",
        "Данные скачаны для того, чтобы мы имели возможность обучить модель на большом объёме данных, добившись более точного результата.\n",
        "\n",
        "К сожалению, времени в рамках хакатона не достаточно, чтобы детально исследовать данные и сделать выборку более сбалансированной. После прохождения во второй этап планируется посвятить время глубокому изучению данных и создать оптимальную для обучения качественной модели выборку."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "p-GxjuPQXAV3",
        "colab_type": "text"
      },
      "source": [
        "# Установка и загрузка библиотек"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "T6S4ituIqfXp",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!pip install pytorch-tabnet transformers"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GkwuN3JXqyt9",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import pandas as pd\n",
        "import shutil\n",
        "import os\n",
        "import numpy as np\n",
        "from transformers import BertTokenizer\n",
        "from transformers import BertModel\n",
        "import torch\n",
        "import random\n",
        "import pickle\n",
        "\n",
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn import preprocessing\n",
        "\n",
        "import matplotlib.pyplot as plt\n",
        "import seaborn as sns"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "p_eN5-YsXavH",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "9c2801b2-d114-41df-e1ca-c5579bfb7634"
      },
      "source": [
        "plt.figure(figsize=(12, 6), dpi=100)\n",
        "sns.set(rc={'figure.figsize':(12, 6), 'figure.dpi':100})\n",
        "plt.style.use('bmh')\n",
        "\n",
        "device = 'cuda' if torch.cuda.is_available() else 'cpu'\n",
        "device = torch.device(device)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1200x600 with 0 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-HdCPfLUXoyZ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "tokenizer = BertTokenizer.from_pretrained('DeepPavlov/rubert-base-cased')\n",
        "bert = BertModel.from_pretrained('DeepPavlov/rubert-base-cased')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qa2349uLXqUt",
        "colab_type": "text"
      },
      "source": [
        "# Датафрейм"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AisT6cVcXHME",
        "colab_type": "text"
      },
      "source": [
        "Перекинем скачанные данные для работы с ними"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wHpTFEGUqxAF",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "!unzip '/content/drive/My Drive/colab_data/hh_salary.zip' -d '/content/raw_data'"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "G2JWbDonX0Z5",
        "colab_type": "text"
      },
      "source": [
        "Создадим из скачанных данных датафрейм для удобства работы:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aTl-LXJrrDKC",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "467dcb56-13fa-4f1b-f11a-b1141d1e23f2"
      },
      "source": [
        "path = 'raw_data'\n",
        "raw_files = [os.path.join(path, csv_file) for csv_file in os.listdir(path) if 'csv' in csv_file]\n",
        "len(raw_files)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "2878"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jLtv4XVvq1Vt",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 513
        },
        "outputId": "afaa85c6-5981-4680-f664-302af01498dc"
      },
      "source": [
        "df = pd.concat([pd.read_csv(csv_file) for csv_file in raw_files], axis=0)\n",
        "df.head(5)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>name</th>\n",
              "      <th>area_id</th>\n",
              "      <th>area_name</th>\n",
              "      <th>salary_from</th>\n",
              "      <th>salary_to</th>\n",
              "      <th>experience</th>\n",
              "      <th>schedule</th>\n",
              "      <th>employment</th>\n",
              "      <th>description</th>\n",
              "      <th>employer</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Инженер-геодезист</td>\n",
              "      <td>95</td>\n",
              "      <td>Тюмень</td>\n",
              "      <td>25000.0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>noExperience</td>\n",
              "      <td>fullDay</td>\n",
              "      <td>full</td>\n",
              "      <td>Обязанности: выполнение инженерно-геодезически...</td>\n",
              "      <td>2671264</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>Ведущий инженер-программист</td>\n",
              "      <td>2</td>\n",
              "      <td>Санкт-Петербург</td>\n",
              "      <td>65000.0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>between1And3</td>\n",
              "      <td>fullDay</td>\n",
              "      <td>full</td>\n",
              "      <td>Обязанности: Разработка управляющих программ в...</td>\n",
              "      <td>46539</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>Менеджер по работе с клиентами (Дмитров, ул. М...</td>\n",
              "      <td>2023</td>\n",
              "      <td>Дмитров</td>\n",
              "      <td>25000.0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>noExperience</td>\n",
              "      <td>shift</td>\n",
              "      <td>full</td>\n",
              "      <td>Компания «ВАЙЛДБЕРРИЗ» – это один из крупнейши...</td>\n",
              "      <td>87021</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>Автоэлектрик-диагност</td>\n",
              "      <td>53</td>\n",
              "      <td>Краснодар</td>\n",
              "      <td>30000.0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>between1And3</td>\n",
              "      <td>shift</td>\n",
              "      <td>full</td>\n",
              "      <td>Обязанности: - проведение компьютерной диагнос...</td>\n",
              "      <td>2681079</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>Продавец-консультант в салон связи Связной</td>\n",
              "      <td>1849</td>\n",
              "      <td>Калач</td>\n",
              "      <td>31000.0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>noExperience</td>\n",
              "      <td>fullDay</td>\n",
              "      <td>full</td>\n",
              "      <td>Ты заряжен энергией и хочешь превратить ее в д...</td>\n",
              "      <td>533</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                                                name  ...  employer\n",
              "0                                  Инженер-геодезист  ...   2671264\n",
              "1                        Ведущий инженер-программист  ...     46539\n",
              "2  Менеджер по работе с клиентами (Дмитров, ул. М...  ...     87021\n",
              "3                              Автоэлектрик-диагност  ...   2681079\n",
              "4         Продавец-консультант в салон связи Связной  ...       533\n",
              "\n",
              "[5 rows x 10 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 20
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "C1dV6SGHRZx8",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 306
        },
        "outputId": "e17d1662-ca45-4529-c914-9535799d1444"
      },
      "source": [
        "df.info()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "<class 'pandas.core.frame.DataFrame'>\n",
            "Int64Index: 1976131 entries, 0 to 985\n",
            "Data columns (total 10 columns):\n",
            " #   Column       Dtype  \n",
            "---  ------       -----  \n",
            " 0   name         object \n",
            " 1   area_id      int64  \n",
            " 2   area_name    object \n",
            " 3   salary_from  float64\n",
            " 4   salary_to    float64\n",
            " 5   experience   object \n",
            " 6   schedule     object \n",
            " 7   employment   object \n",
            " 8   description  object \n",
            " 9   employer     int64  \n",
            "dtypes: float64(2), int64(2), object(6)\n",
            "memory usage: 165.8+ MB\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ObIVr7JlYCv1",
        "colab_type": "text"
      },
      "source": [
        "В наших данных очень скошенная выборка по городам. Получается очень длинный хвост. Поэтому будет достаточно трудно сделать точный предикт зарплат для всех городов России. Вот пример топ-20 городов нашей выборки:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "e0BTVV_D72gm",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 658
        },
        "outputId": "30901239-282b-4a85-84c9-bbf531cfc896"
      },
      "source": [
        "df.area_name.value_counts()[0:20].plot.bar()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.axes._subplots.AxesSubplot at 0x7fbb35b40b70>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 22
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1200x600 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FiM-Upx1YeY6",
        "colab_type": "text"
      },
      "source": [
        "В датасете содержатся различные населённые пункты, при этом имеются и города не из России. Поэтому ограничим нашу выборку лишь несколькими городами России."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xL9QPVZHq7fO",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "cities = ['Санкт-Петербург', 'Москва', 'Екатеринбург', 'Новосибирск', 'Краснодар', 'Казань', 'Нижний Новгород', 'Уфа', 'Красноярск', 'Воронеж', 'Ростов-на-Дону', 'Самара', 'Челябинск', 'Пермь', 'Омск', 'Иркутск', 'Тюмень', 'Волгоград', 'Барнаул', 'Саратов', 'Ярославль', 'Ижевск', 'Тула', 'Томск', 'Сочи', 'Хабаровск', 'Тверь', 'Владивосток', 'Рязань', 'Кемерово', 'Набережные Челны', 'Оренбург', 'Тольятти', 'Белгород', 'Киров (Кировская область)', 'Владимир', 'Ульяновск', 'Липецк', 'Чебоксары', 'Калининград', 'Пенза', 'Новокузнецк', 'Иваново (Ивановская область)', 'Калуга', 'Ставрополь', 'Брянск', 'Сургут', 'Химки', 'Курск', 'Смоленск', 'Вологда', 'Симферополь', 'Подольск (Московская область)', 'Орел', 'Астрахань', 'Караганда', 'Кострома', 'Тамбов', 'Мытищи', 'Улан-Удэ', 'Брест', 'Архангельск', 'Одинцово', 'Мурманск', 'Магнитогорск', 'Зеленоград', 'Новороссийск', 'Люберцы', 'Балашиха', 'Череповец', 'Йошкар-Ола', 'Курган', 'Могилев', 'Гродно', 'Саранск', 'Красногорск', 'Чита', 'Севастополь', 'Домодедово', 'Благовещенск (Амурская область)', 'Псков', 'Нижневартовск', 'Великий Новгород', 'Петрозаводск', 'Волжский(Волгоградская область)', 'Стерлитамак', 'Нижний Тагил', 'Дзержинск (Нижегородская область)', 'Абакан', 'Пятигорск', 'Обнинск', 'Новый Уренгой', 'Салават', 'Королев', 'Таганрог', 'Старый Оскол', 'Энгельс', 'Ангарск', 'Бийск', 'Лобня', 'Южно-Сахалинск', 'Сыктывкар', 'Адлер', 'Пушкино (Московская область)', 'Видное', 'Щелково', 'Дмитров', 'Долгопрудный', 'Нижнекамск', 'Анапа', 'Раменское', 'Братск', 'Ногинск', 'Шахты', 'Петропавловск-Камчатский', 'Альметьевск', 'Солнечногорск', 'Новочеркасск', 'Серпухов', 'Нефтекамск', 'Новомосковск (Тульская область)', 'Миасс', 'Коломна', 'Чехов', 'Ханты-Мансийск', 'Якутск', 'Комсомольск-на-Амуре', 'Армавир', 'Волгодонск', 'Батайск', 'Жуковский', 'Истра', 'Сергиев Посад', 'Орск', 'Рыбинск', 'Реутов', 'Ковров', 'Тобольск', 'Наро-Фоминск', 'Геленджик', 'Клин', 'Электросталь', 'Махачкала', 'Ялта', 'Сызрань', 'Октябрьский (Республика Башкортостан)', 'Каменск-Уральский', 'Березники', 'Балаково', 'Колпино', 'Кстово', 'Всеволожск', 'Аксай (Ростовская область)', 'Магадан', 'Гатчина', 'Орехово-Зуево', 'Троицк (Московская область)', 'Златоуст', 'Елабуга', 'Первоуральск', 'Ачинск', 'Нефтеюганск', 'Муром', 'Пушкин', 'Северодвинск', 'Ноябрьск', 'Уссурийск', 'Ессентуки', 'Елец', 'Азов', 'Верхняя Пышма', 'Прокопьевск', 'Бердск', 'Березовский (Свердловская область)', 'Белово', 'Димитровград', 'Майкоп', 'Выборг', 'Норильск', 'Воскресенск', 'Петропавловск', 'Ухта', 'Находка', 'Ступино', 'Арзамас', 'Великие Луки', 'Котельники', 'Дубна', 'Новочебоксарск', 'Минеральные Воды', 'Туапсе', 'Кингисепп', 'Щербинка', 'Климовск (Московская область)', 'Зеленодольск (Республика Татарстан)', 'Глазов', 'Ленинск-Кузнецкий', 'Сарапул', 'Железногорск (Курская область)', 'Владикавказ', 'Горно-Алтайск', 'Фрязино', 'Воткинск', 'Новокуйбышевск', 'Каменск-Шахтинский', 'Бузулук', 'Невинномысск', 'Копейск', 'Новоалтайск', 'Нахабино', 'Павловский Посад', 'Усинск', 'Керчь', 'Борисоглебск', 'Соликамск', 'Звенигород', 'Егорьевск', 'Александров', 'Славянск-на-Кубани', 'Кисловодск', 'Ржев', 'Большой Камень', 'Камышин', 'Тихорецк', 'Можайск', 'Элиста', 'Темрюк', 'Вязьма', 'Когалым', 'Бугульма', 'Ейск', 'Жигулевск', 'Асбест', 'Нальчик', 'Черкесск', 'Биробиджан', 'Тихвин', 'Сестрорецк', 'Юрга', 'Волоколамск', 'Ростов (Ярославская область)', 'Балабаново', 'Сальск', 'Петергоф', 'Гусь-Хрустальный', 'Ломоносов', 'Евпатория', 'Волжск', 'Грязи', 'Краснознаменск (Московская область)', 'Салехард', 'Алушта', 'Нерюнгри', 'Тутаев', 'Усть-Илимск', 'Шадринск', 'Курчатов', 'Воркута', 'Новотроицк (Оренбургская область)', 'Гагарин', 'Минусинск', 'Кинешма', 'Благовещенск (Республика Башкортостан)', 'Ахтубинск', 'Краснокамск', 'Лесосибирск', 'Чапаевск', 'Озерск (Челябинская область)', 'Мирный (Республика Саха (Якутия))', 'Кронштадт', 'Киров (Калужская область)', 'Донецк (Ростовская область)', 'Березовский (Кемеровская область)', 'Тамань', 'Грозный', 'Строитель (Белгородская область)', 'Суздаль', 'Зеленогорск (Ленинградская обл)', 'Шлиссельбург', 'Быково (Московская область)', 'Амурск', 'Лазаревское', 'Котлас', 'Калач-на-Дону', 'Североморск', 'Зерноград', 'Отрадный (Самарская область)', 'Котовск (Тамбовская область)', 'Урюпинск', 'Заречный (Свердловская область)', 'Радужный (Ханты-Мансийский АО - Югра)', 'Заречный (Пензенская область)', 'Красноармейск (Московская область)', 'Черноголовка', 'Павловск (Ленинградская область)', 'Черногорск', 'Тында', 'Бодайбо', 'Ленинградская станица', 'Фролово', 'Рощино (Ленинградская область)', 'Лермонтов', 'Славгород (Алтайский край)', 'Николаевск-на-Амуре', 'Валдай', 'Саянск', 'Приморско-Ахтарск', 'Гурьевск (Кемеровская область)', 'Горки-10 (Московская область, Одинцовский район)', 'Дубна (Тульская область)', 'Гурьевск (Калининградская область)', 'Забайкальск', 'Запрудня (Московская область)', 'Каспийск', 'Александровск (Пермский край)', 'Славянка (Приморский край)', 'Павловск (Алтайский край)', 'Иннополис', 'Нарьян-Мар', 'Шексна', 'Печоры', 'Нижнеудинск', 'Дно', 'Краснослободск (Волгоградская область)', 'Советский (Ханты-Мансийский АО)', 'Советск (Калининградская область)', 'Красный Бор (Ленинградская область)', 'Любань (Ленинградская область)', 'Аннино (Ленинградская область)', 'Быков (Сахалинская область)', 'Птичник (Еврейская АО)']"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "39np_2eVZCTj",
        "colab_type": "text"
      },
      "source": [
        "Выберем только те значения нашего датафрейма, которые относятся к выбранным нами городам.\n",
        "\n",
        "Также удалим пропуски и перемешаем данные для более гладкого процесса обучения модели.\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8eICGWZlP0vg",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "pattern = '|'.join(cities)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gwtPU39xSAIY",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 292
        },
        "outputId": "68cc9797-4f88-477c-952b-e684173e5ba1"
      },
      "source": [
        "df = df[df.area_name.str.contains(pattern) == True]\n",
        "df = df[['name', 'area_name', 'salary_from', 'experience', 'schedule', 'employment']]\n",
        "df = df.dropna()\n",
        "df = df.sample(frac=1).reset_index(drop=True)\n",
        "df.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/pandas/core/strings.py:1954: UserWarning: This pattern has match groups. To actually get the groups, use str.extract.\n",
            "  return func(self, *args, **kwargs)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>name</th>\n",
              "      <th>area_name</th>\n",
              "      <th>salary_from</th>\n",
              "      <th>experience</th>\n",
              "      <th>schedule</th>\n",
              "      <th>employment</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Администратор (Самара, Московское шоссе, 29)</td>\n",
              "      <td>Самара</td>\n",
              "      <td>25000.0</td>\n",
              "      <td>between1And3</td>\n",
              "      <td>fullDay</td>\n",
              "      <td>full</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>Продавец-кассир (Москва, ул Гришина, 9)</td>\n",
              "      <td>Москва</td>\n",
              "      <td>33500.0</td>\n",
              "      <td>noExperience</td>\n",
              "      <td>fullDay</td>\n",
              "      <td>full</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>Повар</td>\n",
              "      <td>Москва</td>\n",
              "      <td>29000.0</td>\n",
              "      <td>between1And3</td>\n",
              "      <td>shift</td>\n",
              "      <td>full</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>Специалист по документообороту</td>\n",
              "      <td>Нижний Новгород</td>\n",
              "      <td>30000.0</td>\n",
              "      <td>between1And3</td>\n",
              "      <td>fullDay</td>\n",
              "      <td>full</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>Оператор</td>\n",
              "      <td>Тамбов</td>\n",
              "      <td>20000.0</td>\n",
              "      <td>between1And3</td>\n",
              "      <td>shift</td>\n",
              "      <td>full</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                                           name  ... employment\n",
              "0  Администратор (Самара, Московское шоссе, 29)  ...       full\n",
              "1       Продавец-кассир (Москва, ул Гришина, 9)  ...       full\n",
              "2                                         Повар  ...       full\n",
              "3                Специалист по документообороту  ...       full\n",
              "4                                      Оператор  ...       full\n",
              "\n",
              "[5 rows x 6 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 25
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zlSLTy4gZPxq",
        "colab_type": "text"
      },
      "source": [
        "В названиях профессий иногда встречается лишняя информация. Например, иногда в скобках указывают регион размещения вакансии. Нам же нужно оставить только важную для данного признака информацию, поэтому немного почистим признак `names`."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tiN-QxphSbE-",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "df.name = df.name.str.replace(r'\\(.*?\\)', '')\n",
        "df.name = df.name.str.replace('/', '')\n",
        "df.name = df.name.str.replace(',', '')\n",
        "df.name = df.name.str.replace(' +', ' ')\n",
        "df.name = df.name.str.strip()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wxgVIReAZkAe",
        "colab_type": "text"
      },
      "source": [
        "Как видим, данных в ходе обработки стало немного меньше, но теперь они более качественны."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wdaOgSDcTc1J",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 238
        },
        "outputId": "7d4b3f8f-eb25-4ebf-9ba9-a9901db38c98"
      },
      "source": [
        "df.info()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "<class 'pandas.core.frame.DataFrame'>\n",
            "RangeIndex: 1373240 entries, 0 to 1373239\n",
            "Data columns (total 6 columns):\n",
            " #   Column       Non-Null Count    Dtype  \n",
            "---  ------       --------------    -----  \n",
            " 0   name         1373240 non-null  object \n",
            " 1   area_name    1373240 non-null  object \n",
            " 2   salary_from  1373240 non-null  float64\n",
            " 3   experience   1373240 non-null  object \n",
            " 4   schedule     1373240 non-null  object \n",
            " 5   employment   1373240 non-null  object \n",
            "dtypes: float64(1), object(5)\n",
            "memory usage: 62.9+ MB\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "h8YdtmY6ZtsL",
        "colab_type": "text"
      },
      "source": [
        "# Преобразование датафрейма в вектора\n",
        "\n",
        "Просто так, сырые данные мы не можем подавать нашей модели на обучение. Поэтому преобразуем каждый параметр в нужный вектор."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6xQ2K2HAaA_4",
        "colab_type": "text"
      },
      "source": [
        "Первым подвергнется преобразованию параметр `names` - название вакансии.\n",
        "\n",
        "существует множество способов векторизации текстовой информации. От самых простых, типа \"мешка слов\" и TF-IDF, до более сложных, типа word2vec и GloVe.\n",
        "\n",
        "Мы будем векторизовать названия вакансий с помощью предобученного на русских текстах Берта - **RuBert** от **DeepPavlov**. Это, конечно, более длительный и более ресурсоёмкий процесс, нежели, например, \"мешок слов\", но более эффективный. Вот некоторые его преимущества:\n",
        "\n",
        "* Избавляемся от проблемы _Out of Vocabulary_ на инференсе. Т.е. мы сможем обрабатывать даже те слова, которые модель не видела в ходе обучения.\n",
        "* Разрешаем проблему синонимов и омонимов. Так как Берт - модель языка, мы получаем не просто вектора слов, но вектора со смыслами слов, и наша модель, получая вектора от Берта, будет понимать, что, например, такие вакансии, как \"уборщик\" и \"уборщица\" следует объединять вместе.\n",
        "* Разрешаем проблему опечаток и неправильных написаний слов. (После победы и переходу к следующему этапу планируется провести эксперимент, взяв не RuBert, а Bert Base от Google. Как нам теперь видится, эта языковая модель должна лучше разрешать проблему опечаток и неправильного написания слов).\n",
        "\n",
        "Для получения смысла выходной последовательности часто берут первый или последний токен с выходного слоя Берта (пулинг). Мы же поступим иначе. Как советуют разработчики группы **HuggingFace**, лучше брать векторы всех токенов на выходе Берта и усреднять их. Поэтому мы, чтобы извлечь максимум информации из данного признака, берём у Берта вектор каждого токена и усредняем его.\n",
        "\n",
        "Чтобы ускорить процесс конвертации, лучше применять GPU.\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3G24Ze0tUpGT",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "%%time\n",
        "\n",
        "bert.to(device)\n",
        "bert.eval();\n",
        "\n",
        "step = 10_000\n",
        "names_vectors = False\n",
        "\n",
        "for batch_i in range(0, len(df), step):\n",
        "    batch = df.name[batch_i:batch_i+step]\n",
        "    inputs = tokenizer(\n",
        "        batch.to_list(),\n",
        "        padding='max_length',\n",
        "        truncation=True,\n",
        "        max_length=20,\n",
        "        return_tensors='pt',\n",
        "        return_attention_mask=True,\n",
        "    )\n",
        "    inputs.to(device)\n",
        "    with torch.no_grad():\n",
        "        output = bert(**inputs)\n",
        "    total_sum = torch.zeros_like(output[0][:, 1, :])\n",
        "    for i in range(0, output[0].shape[1]):\n",
        "        total_sum += output[0][:, i, :]\n",
        "    total_sum = total_sum / output[0].shape[1]\n",
        "    # names_vectors.extend(total_sum.detach().cpu().numpy())\n",
        "    total_sum = total_sum.detach().cpu().numpy()\n",
        "    if names_vectors is False:\n",
        "        names_vectors = total_sum\n",
        "    else:\n",
        "        names_vectors = np.vstack((names_vectors, total_sum))\n",
        "    \n",
        "    if batch_i % 100_000 == 0:\n",
        "        print(batch_i)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Af12pH5melQf",
        "colab_type": "text"
      },
      "source": [
        "Проверим, все ли признаки сконвертировались:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-BLChYEmepNA",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "b8eb8b71-abf4-4dd3-84b9-b92593827ddb"
      },
      "source": [
        "names_vectors.shape, df.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "((1373240, 768), (1373240, 6))"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 16
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6IXVXWY8esrD",
        "colab_type": "text"
      },
      "source": [
        "Великолепно! Теперь сохраним наши вектора на диск, чтобы освободить память, и приступим к преобразованию других признаков."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XVCtmJGCe2gl",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "np.save('/content/drive/My Drive/colab_data/salary_tabnetL_data/names.npy', names_vectors)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CapaA4F4d0On",
        "colab_type": "text"
      },
      "source": [
        "# Признаки в индексы\n",
        "\n",
        "Теперь преобразуем другие признаки. Мы создадим из текстовых признаков их индексы. Эти индексы будем подавать модели, а она из них самостоятельно будет получать эмбеддинги - плотные вектора этих признаков.\n",
        "\n",
        "Эти преобразования намного быстрее и легче, их можно выполнять на CPU."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nvnIY_dHU2Lm",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "experience = set()\n",
        "schedule = set()\n",
        "employment = set()\n",
        "city = set()\n",
        "\n",
        "experience.update(df.experience.unique().tolist())\n",
        "schedule.update(df.schedule.unique().tolist())\n",
        "employment.update(df.employment.unique().tolist())\n",
        "city.update(df.area_name.unique().tolist())"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lBLVOsWhgs2J",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "experience2id = {name:id for id, name in enumerate(experience)}\n",
        "schedule2id = {name:id for id, name in enumerate(schedule)}\n",
        "employment2id = {name:id for id, name in enumerate(employment)}\n",
        "city2id = {name:id for id, name in enumerate(city)}"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xNbNdNMRg4no",
        "colab_type": "text"
      },
      "source": [
        "Пример индекса:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "80nFpHcSg1ss",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "5c25901a-ae54-4187-ad7c-b369a0ef499e"
      },
      "source": [
        "experience2id"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "{'between1And3': 1, 'between3And6': 3, 'moreThan6': 2, 'noExperience': 0}"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 36
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Z2oiyU4UhDet",
        "colab_type": "text"
      },
      "source": [
        "Сохраним все индексы. Они нам понадобятся на инференсе модели."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZC56qYh6hAws",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "with open('/content/drive/My Drive/colab_data/salary_tabnetL_data/experience2id.pickle', 'wb') as handle:\n",
        "    pickle.dump(experience2id, handle, protocol=pickle.HIGHEST_PROTOCOL)\n",
        "with open('/content/drive/My Drive/colab_data/salary_tabnetL_data/schedule2id.pickle', 'wb') as handle:\n",
        "    pickle.dump(schedule2id, handle, protocol=pickle.HIGHEST_PROTOCOL)\n",
        "with open('/content/drive/My Drive/colab_data/salary_tabnetL_data/employment2id.pickle', 'wb') as handle:\n",
        "    pickle.dump(employment2id, handle, protocol=pickle.HIGHEST_PROTOCOL)\n",
        "with open('/content/drive/My Drive/colab_data/salary_tabnetL_data/city2id.pickle', 'wb') as handle:\n",
        "    pickle.dump(city2id, handle, protocol=pickle.HIGHEST_PROTOCOL)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZfFPSWfPhbF7",
        "colab_type": "text"
      },
      "source": [
        "Теперь преобразуем все признаки нашего датасета в индексы. И заодно сохраним отдельно нашу целевую метку - `salary_from`."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "r0SYKt9bg6LO",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "73036ce8-d82b-4ba4-b20c-c309858d84d2"
      },
      "source": [
        "%%time\n",
        "experience_label = []\n",
        "schedule_label = []\n",
        "employment_label = []\n",
        "city_label = []\n",
        "salary_from\t= []\n",
        "\n",
        "for index in df.index:\n",
        "    experience_label.append(experience2id[df.experience.iloc[index]])\n",
        "    schedule_label.append(schedule2id[df.schedule.iloc[index]])\n",
        "    employment_label.append(employment2id[df.employment.iloc[index]])\n",
        "    city_label.append(city2id[df.area_name.iloc[index]])\n",
        "    salary_from.append(df.salary_from.iloc[index])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "CPU times: user 1min 57s, sys: 71.2 ms, total: 1min 57s\n",
            "Wall time: 1min 57s\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GxfmCv4digTL",
        "colab_type": "text"
      },
      "source": [
        "Стандартизируем нашу целевую метку для получения более качественного результата."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wCQG2cAvhwVF",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "salary_from = np.array(salary_from)\n",
        "salary_from_scaled = preprocessing.scale(salary_from)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9FE9hw7Diw4k",
        "colab_type": "text"
      },
      "source": [
        "Пример стандартизированного и обычного значения зарплаты:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YNSG3NIwh605",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "2479302b-76ca-4610-d595-ee88a156f635"
      },
      "source": [
        "salary_from_scaled[22], salary_from[22]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(-0.8137864340371029, 20000.0)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 40
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wdjYJ07bi8Bn",
        "colab_type": "text"
      },
      "source": [
        "Сохраним стандартизированное значение таргета, мы его будем подавать модели в ходе обучения, и обычное значение - оно нам понадобится для того, чтобы из стандартизированного предикта получить \"человекочитаемую\" цифру."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ScDEoJ-Xi1J2",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "np.save('/content/drive/My Drive/colab_data/salary_tabnetL_data/salary_scaled.npy', salary_from_scaled)\n",
        "np.save('/content/drive/My Drive/colab_data/salary_tabnetL_data/salary_orig.npy', salary_from)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BIU71qPPjlM4",
        "colab_type": "text"
      },
      "source": [
        "Преобразуем наши признаки из типа `list` в тип `numpy.array`, приведём их к нужной размерности и сохраним."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "r4eT6bpkjkNn",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "experience_label = np.array(experience_label)\n",
        "schedule_label = np.array(schedule_label)\n",
        "employment_label = np.array(employment_label)\n",
        "city_label = np.array(city_label)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sKDj4jf5jabe",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "experience_label = np.reshape(experience_label, (experience_label.shape[0], 1))\n",
        "schedule_label = np.reshape(schedule_label, (schedule_label.shape[0], 1))\n",
        "employment_label = np.reshape(employment_label, (employment_label.shape[0], 1))\n",
        "city_label = np.reshape(city_label, (city_label.shape[0], 1))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bdoY2BbIkCe2",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "np.save('/content/drive/My Drive/colab_data/salary_tabnetL_data/experience_label.npy', experience_label)\n",
        "np.save('/content/drive/My Drive/colab_data/salary_tabnetL_data/schedule_label.npy', schedule_label)\n",
        "np.save('/content/drive/My Drive/colab_data/salary_tabnetL_data/employment_label.npy', employment_label)\n",
        "np.save('/content/drive/My Drive/colab_data/salary_tabnetL_data/city_label.npy', city_label)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hDQyGPKukgLi",
        "colab_type": "text"
      },
      "source": [
        "Весь датасет подготовлен. Теперь можно приступить к обучению модели."
      ]
    }
  ]
}