MII_Salin_Oleg_PIbd-33/lec2unicorns.ipynb
2024-10-22 18:54:39 +04:00

724 lines
23 KiB
Plaintext
Raw Permalink Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Загрузка данных в DataFrame"
]
},
{
"cell_type": "code",
"execution_count": 76,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"<class 'pandas.core.frame.DataFrame'>\n",
"Index: 100 entries, Bytedance to iCapital Network\n",
"Data columns (total 9 columns):\n",
" # Column Non-Null Count Dtype \n",
"--- ------ -------------- ----- \n",
" 0 Valuation 100 non-null object\n",
" 1 Country 100 non-null object\n",
" 2 State 79 non-null object\n",
" 3 City 99 non-null object\n",
" 4 Industries 99 non-null object\n",
" 5 FoundedYear 100 non-null int64 \n",
" 6 Name of Founders 100 non-null object\n",
" 7 TotalFunding 100 non-null object\n",
" 8 Number of Employees 100 non-null object\n",
"dtypes: int64(1), object(8)\n",
"memory usage: 7.8+ KB\n",
"(100, 10)\n"
]
},
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>Valuation</th>\n",
" <th>Country</th>\n",
" <th>State</th>\n",
" <th>City</th>\n",
" <th>Industries</th>\n",
" <th>FoundedYear</th>\n",
" <th>Name of Founders</th>\n",
" <th>TotalFunding</th>\n",
" <th>Number of Employees</th>\n",
" <th>IsChina</th>\n",
" </tr>\n",
" <tr>\n",
" <th>Company</th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" <th></th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>Bytedance</th>\n",
" <td>140.0</td>\n",
" <td>China</td>\n",
" <td>Beijing</td>\n",
" <td>Beijing</td>\n",
" <td>Content, Data Mining, Internet</td>\n",
" <td>2012</td>\n",
" <td>Yiming Zhang</td>\n",
" <td>7440.00</td>\n",
" <td>10.000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>SpaceX</th>\n",
" <td>100.3</td>\n",
" <td>United States</td>\n",
" <td>California</td>\n",
" <td>Hawthorne</td>\n",
" <td>Aerospace, Manufacturing, Space Travel, Transp...</td>\n",
" <td>2002</td>\n",
" <td>Elon Musk</td>\n",
" <td>383.02</td>\n",
" <td>5,000-10,000</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>Stripe</th>\n",
" <td>95.0</td>\n",
" <td>United States</td>\n",
" <td>California</td>\n",
" <td>San Francisco</td>\n",
" <td>Finance, FinTech, Mobile Payments, SaaS</td>\n",
" <td>2010</td>\n",
" <td>John Collison, Patrick Collison</td>\n",
" <td>300.00</td>\n",
" <td>1,000-5,000</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>Klarna</th>\n",
" <td>45.6</td>\n",
" <td>Sweden</td>\n",
" <td>NaN</td>\n",
" <td>Stockholm</td>\n",
" <td>E-Commerce, FinTech, Payments, Shopping</td>\n",
" <td>2005</td>\n",
" <td>Niklas Adalberth, Sebastian Siemiatkowski, Vic...</td>\n",
" <td>3471.72</td>\n",
" <td>5,000-10,000</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>Epic Games</th>\n",
" <td>42.0</td>\n",
" <td>United States</td>\n",
" <td>North Carolina</td>\n",
" <td>Cary</td>\n",
" <td>Developer Platform, Gaming, Software, Video Games</td>\n",
" <td>1991</td>\n",
" <td>Mark Rein, Tim Sweeney</td>\n",
" <td>544.93</td>\n",
" <td>1,000-5,000</td>\n",
" <td>0</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"</div>"
],
"text/plain": [
" Valuation Country State City \\\n",
"Company \n",
"Bytedance 140.0 China Beijing Beijing \n",
"SpaceX 100.3 United States California Hawthorne \n",
"Stripe 95.0 United States California San Francisco \n",
"Klarna 45.6 Sweden NaN Stockholm \n",
"Epic Games 42.0 United States North Carolina Cary \n",
"\n",
" Industries FoundedYear \\\n",
"Company \n",
"Bytedance Content, Data Mining, Internet 2012 \n",
"SpaceX Aerospace, Manufacturing, Space Travel, Transp... 2002 \n",
"Stripe Finance, FinTech, Mobile Payments, SaaS 2010 \n",
"Klarna E-Commerce, FinTech, Payments, Shopping 2005 \n",
"Epic Games Developer Platform, Gaming, Software, Video Games 1991 \n",
"\n",
" Name of Founders TotalFunding \\\n",
"Company \n",
"Bytedance Yiming Zhang 7440.00 \n",
"SpaceX Elon Musk 383.02 \n",
"Stripe John Collison, Patrick Collison 300.00 \n",
"Klarna Niklas Adalberth, Sebastian Siemiatkowski, Vic... 3471.72 \n",
"Epic Games Mark Rein, Tim Sweeney 544.93 \n",
"\n",
" Number of Employees IsChina \n",
"Company \n",
"Bytedance 10.000 1 \n",
"SpaceX 5,000-10,000 0 \n",
"Stripe 1,000-5,000 0 \n",
"Klarna 5,000-10,000 0 \n",
"Epic Games 1,000-5,000 0 "
]
},
"execution_count": 76,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"import pandas as pd\n",
"\n",
"df = pd.read_csv(\"data/unicorns.csv\", index_col=\"Company\", sep=';')\n",
"\n",
"df.info()\n",
"\n",
"df[\"Valuation\"] = df[\"Valuation\"].apply(\n",
" lambda x: float(x[:-4].replace(',', '.')),\n",
")\n",
"\n",
"df[\"TotalFunding\"] = df[\"TotalFunding\"].apply(\n",
" lambda x: float(x.strip(\"$M\").replace(\",\", \"\")),\n",
")\n",
"\n",
"df[\"IsChina\"] = [int(country == 'China') for country in df[\"Country\"]]\n",
"print(df.shape)\n",
"\n",
"df.head()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Получение сведений о пропущенных данных"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Типы пропущенных данных:\n",
"- None - представление пустых данных в Python\n",
"- NaN - представление пустых данных в Pandas\n",
"- '' - пустая строка"
]
},
{
"cell_type": "code",
"execution_count": 77,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Valuation 0\n",
"Country 0\n",
"State 21\n",
"City 1\n",
"Industries 1\n",
"FoundedYear 0\n",
"Name of Founders 0\n",
"TotalFunding 0\n",
"Number of Employees 0\n",
"IsChina 0\n",
"dtype: int64\n",
"\n",
"Valuation False\n",
"Country False\n",
"State True\n",
"City True\n",
"Industries True\n",
"FoundedYear False\n",
"Name of Founders False\n",
"TotalFunding False\n",
"Number of Employees False\n",
"IsChina False\n",
"dtype: bool\n",
"\n",
"State процент пустых значений: %21.00\n",
"City процент пустых значений: %1.00\n",
"Industries процент пустых значений: %1.00\n"
]
}
],
"source": [
"# Количество пустых значений признаков\n",
"print(df.isnull().sum())\n",
"\n",
"print()\n",
"\n",
"# Есть ли пустые значения признаков\n",
"print(df.isnull().any())\n",
"\n",
"print()\n",
"\n",
"# Процент пустых значений признаков\n",
"for i in df.columns:\n",
" null_rate = df[i].isnull().sum() / len(df) * 100\n",
" if null_rate > 0:\n",
" print(f\"{i} процент пустых значений: %{null_rate:.2f}\")"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Заполнение пропущенных данных\n",
"\n",
"https://pythonmldaily.com/posts/pandas-dataframes-search-drop-empty-values\n",
"\n",
"https://scales.arabpsychology.com/stats/how-to-fill-nan-values-with-median-in-pandas/"
]
},
{
"cell_type": "code",
"execution_count": 78,
"metadata": {},
"outputs": [],
"source": [
"# fillna_df = df.fillna(0)\n",
"\n",
"# print(fillna_df.shape)\n",
"\n",
"# print(fillna_df.isnull().any())\n",
"\n",
"# # Замена пустых данных на 0\n",
"# df[\"AgeFillNA\"] = df[\"Age\"].fillna(0)\n",
"\n",
"# # Замена пустых данных на медиану\n",
"# df[\"AgeFillMedian\"] = df[\"Age\"].fillna(df[\"Age\"].median())\n",
"\n",
"# df.tail()"
]
},
{
"cell_type": "code",
"execution_count": 79,
"metadata": {},
"outputs": [],
"source": [
"# df[\"AgeCopy\"] = df[\"Age\"]\n",
"\n",
"# # Замена данных сразу в DataFrame без копирования\n",
"# df.fillna({\"AgeCopy\": 0}, inplace=True)\n",
"\n",
"# df.tail()"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Удаление наблюдений с пропусками"
]
},
{
"cell_type": "code",
"execution_count": 80,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"(100, 7)\n",
"Valuation False\n",
"Country False\n",
"FoundedYear False\n",
"Name of Founders False\n",
"TotalFunding False\n",
"Number of Employees False\n",
"IsChina False\n",
"dtype: bool\n"
]
}
],
"source": [
"df = df.dropna(axis=1)\n",
"\n",
"print(df.shape)\n",
"\n",
"print(df.isnull().any())"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Создание выборок данных\n",
"\n",
"Библиотека scikit-learn\n",
"\n",
"https://scikit-learn.org/stable/index.html"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"<img src=\"assets/lec2-split.png\" width=\"600\" style=\"background-color: white\">"
]
},
{
"cell_type": "code",
"execution_count": 81,
"metadata": {},
"outputs": [],
"source": [
"# Функция для создания выборок\n",
"from sklearn.model_selection import train_test_split\n",
"\n",
"\n",
"def split_stratified_into_train_val_test(\n",
" df_input,\n",
" stratify_colname=\"y\",\n",
" frac_train=0.6,\n",
" frac_val=0.15,\n",
" frac_test=0.25,\n",
" random_state=None,\n",
"):\n",
" \"\"\"\n",
" Splits a Pandas dataframe into three subsets (train, val, and test)\n",
" following fractional ratios provided by the user, where each subset is\n",
" stratified by the values in a specific column (that is, each subset has\n",
" the same relative frequency of the values in the column). It performs this\n",
" splitting by running train_test_split() twice.\n",
"\n",
" Parameters\n",
" ----------\n",
" df_input : Pandas dataframe\n",
" Input dataframe to be split.\n",
" stratify_colname : str\n",
" The name of the column that will be used for stratification. Usually\n",
" this column would be for the label.\n",
" frac_train : float\n",
" frac_val : float\n",
" frac_test : float\n",
" The ratios with which the dataframe will be split into train, val, and\n",
" test data. The values should be expressed as float fractions and should\n",
" sum to 1.0.\n",
" random_state : int, None, or RandomStateInstance\n",
" Value to be passed to train_test_split().\n",
"\n",
" Returns\n",
" -------\n",
" df_train, df_val, df_test :\n",
" Dataframes containing the three splits.\n",
" \"\"\"\n",
"\n",
" if frac_train + frac_val + frac_test != 1.0:\n",
" raise ValueError(\n",
" \"fractions %f, %f, %f do not add up to 1.0\"\n",
" % (frac_train, frac_val, frac_test)\n",
" )\n",
"\n",
" if stratify_colname not in df_input.columns:\n",
" raise ValueError(\"%s is not a column in the dataframe\" % (stratify_colname))\n",
"\n",
" X = df_input # Contains all columns.\n",
" y = df_input[\n",
" [stratify_colname]\n",
" ] # Dataframe of just the column on which to stratify.\n",
"\n",
" # Split original dataframe into train and temp dataframes.\n",
" df_train, df_temp, y_train, y_temp = train_test_split(\n",
" X, y, stratify=y, test_size=(1.0 - frac_train), random_state=random_state\n",
" )\n",
"\n",
" # Split the temp dataframe into val and test dataframes.\n",
" relative_frac_test = frac_test / (frac_val + frac_test)\n",
" df_val, df_test, y_val, y_test = train_test_split(\n",
" df_temp,\n",
" y_temp,\n",
" stratify=y_temp,\n",
" test_size=relative_frac_test,\n",
" random_state=random_state,\n",
" )\n",
"\n",
" assert len(df_input) == len(df_train) + len(df_val) + len(df_test)\n",
"\n",
" return df_train, df_val, df_test"
]
},
{
"cell_type": "code",
"execution_count": 82,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"IsChina\n",
"0 86\n",
"1 14\n",
"Name: count, dtype: int64\n",
"Обучающая выборка: (60, 3)\n",
"IsChina\n",
"0 52\n",
"1 8\n",
"Name: count, dtype: int64\n",
"Контрольная выборка: (20, 3)\n",
"IsChina\n",
"0 17\n",
"1 3\n",
"Name: count, dtype: int64\n",
"Тестовая выборка: (20, 3)\n",
"IsChina\n",
"0 17\n",
"1 3\n",
"Name: count, dtype: int64\n"
]
}
],
"source": [
"# Вывод распределения количества наблюдений по меткам (классам)\n",
"print(df.IsChina.value_counts())\n",
"\n",
"data = df[[\"TotalFunding\", \"Valuation\", \"IsChina\"]].copy()\n",
"\n",
"df_train, df_val, df_test = split_stratified_into_train_val_test(\n",
" data,\n",
" stratify_colname=\"IsChina\",\n",
" frac_train=0.60,\n",
" frac_val=0.20,\n",
" frac_test=0.20,\n",
")\n",
"\n",
"print(\"Обучающая выборка: \", df_train.shape)\n",
"print(df_train.IsChina.value_counts())\n",
"\n",
"print(\"Контрольная выборка: \", df_val.shape)\n",
"print(df_val.IsChina.value_counts())\n",
"\n",
"print(\"Тестовая выборка: \", df_test.shape)\n",
"print(df_test.IsChina.value_counts())"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Выборка с избытком (oversampling)\n",
"\n",
"https://www.blog.trainindata.com/oversampling-techniques-for-imbalanced-data/\n",
"\n",
"https://datacrayon.com/machine-learning/class-imbalance-and-oversampling/\n",
"\n",
"Выборка с недостатком (undersampling)\n",
"\n",
"https://machinelearningmastery.com/random-oversampling-and-undersampling-for-imbalanced-classification/\n",
"\n",
"Библиотека imbalanced-learn\n",
"\n",
"https://imbalanced-learn.org/stable/"
]
},
{
"cell_type": "code",
"execution_count": 83,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Обучающая выборка: (60, 3)\n",
"IsChina\n",
"0 52\n",
"1 8\n",
"Name: count, dtype: int64\n",
"Обучающая выборка после oversampling: (105, 3)\n",
"IsChina\n",
"1 53\n",
"0 52\n",
"Name: count, dtype: int64\n"
]
},
{
"data": {
"text/html": [
"<div>\n",
"<style scoped>\n",
" .dataframe tbody tr th:only-of-type {\n",
" vertical-align: middle;\n",
" }\n",
"\n",
" .dataframe tbody tr th {\n",
" vertical-align: top;\n",
" }\n",
"\n",
" .dataframe thead th {\n",
" text-align: right;\n",
" }\n",
"</style>\n",
"<table border=\"1\" class=\"dataframe\">\n",
" <thead>\n",
" <tr style=\"text-align: right;\">\n",
" <th></th>\n",
" <th>TotalFunding</th>\n",
" <th>Valuation</th>\n",
" <th>IsChina</th>\n",
" </tr>\n",
" </thead>\n",
" <tbody>\n",
" <tr>\n",
" <th>0</th>\n",
" <td>208.000000</td>\n",
" <td>9.500000</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>1</th>\n",
" <td>4044.200000</td>\n",
" <td>15.500000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>2</th>\n",
" <td>447.120000</td>\n",
" <td>6.500000</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>3</th>\n",
" <td>2121.000000</td>\n",
" <td>6.600000</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>4</th>\n",
" <td>2686.010000</td>\n",
" <td>39.000000</td>\n",
" <td>0</td>\n",
" </tr>\n",
" <tr>\n",
" <th>...</th>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" <td>...</td>\n",
" </tr>\n",
" <tr>\n",
" <th>100</th>\n",
" <td>1306.334794</td>\n",
" <td>14.179790</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>101</th>\n",
" <td>1492.220325</td>\n",
" <td>10.610196</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>102</th>\n",
" <td>1125.438822</td>\n",
" <td>16.887502</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>103</th>\n",
" <td>1728.312129</td>\n",
" <td>7.708914</td>\n",
" <td>1</td>\n",
" </tr>\n",
" <tr>\n",
" <th>104</th>\n",
" <td>1785.708076</td>\n",
" <td>7.004370</td>\n",
" <td>1</td>\n",
" </tr>\n",
" </tbody>\n",
"</table>\n",
"<p>105 rows × 3 columns</p>\n",
"</div>"
],
"text/plain": [
" TotalFunding Valuation IsChina\n",
"0 208.000000 9.500000 0\n",
"1 4044.200000 15.500000 1\n",
"2 447.120000 6.500000 0\n",
"3 2121.000000 6.600000 1\n",
"4 2686.010000 39.000000 0\n",
".. ... ... ...\n",
"100 1306.334794 14.179790 1\n",
"101 1492.220325 10.610196 1\n",
"102 1125.438822 16.887502 1\n",
"103 1728.312129 7.708914 1\n",
"104 1785.708076 7.004370 1\n",
"\n",
"[105 rows x 3 columns]"
]
},
"execution_count": 83,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"from imblearn.over_sampling import ADASYN\n",
"\n",
"ada = ADASYN()\n",
"\n",
"print(\"Обучающая выборка: \", df_train.shape)\n",
"print(df_train.IsChina.value_counts())\n",
"\n",
"X_resampled, y_resampled = ada.fit_resample(df_train, df_train[\"IsChina\"]) # type: ignore\n",
"df_train_adasyn = pd.DataFrame(X_resampled)\n",
"\n",
"print(\"Обучающая выборка после oversampling: \", df_train_adasyn.shape)\n",
"print(df_train_adasyn.IsChina.value_counts())\n",
"\n",
"df_train_adasyn"
]
}
],
"metadata": {
"kernelspec": {
"display_name": ".venv",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.12.4"
}
},
"nbformat": 4,
"nbformat_minor": 2
}