Pipeline

import pandas as pd
from sklearn.model_selection import cross_val_score
from sklearn.preprocessing import OneHotEncoder
from sklearn.preprocessing import StandardScaler
from sklearn.ensemble import RandomForestClassifier
from sklearn.linear_model import LogisticRegression
from sklearn.svm import SVC
from sklearn.impute import SimpleImputer
from sklearn.compose import ColumnTransformer
from sklearn.preprocessing import FunctionTransformer
from sklearn.pipeline import Pipeline
from sklearn.compose import make_column_selector
 
df = pd.read_csv('homework.csv')
 
def filter_data(df):
    columns_to_drop = [
       'id',
       'url',
       'region',
       'region_url',
       'price',
       'manufacturer',
       'image_url',
       'description',
       'posting_date',
       'lat',
       'long']
    return df.drop(columns_to_drop, axis=1)
 
def calculate_outliers(df):
    q25 = df['year'].quantile(0.25)
    q75 = df['year'].quantile(0.75)
    iqr = q75 - q25
    df.loc[df['year'] < (q25 - 1.5 * iqr), 'year'] = round(q25 - 1.5 * iqr)
    df.loc[df['year'] > (q75 + 1.5 * iqr), 'year'] = round(q25 + 1.5 * iqr)
    return df['year']
 
 
def new_predictor(df):
    df.loc[:, 'age_category'] = df['year'].apply(lambda x: 'new' if x > 2013 else ('old' if x < 2006 else 'average'))
    df.loc[:, 'short_model'] = df.apply(lambda x: x.model.split(' ')[0], axis=1)
    return df
 
 
def filter_data2(df):
    columns_to_drop = [
        'year',
        'model',
        'fuel',
        'odometer',
        'title_status',
        'transmission',
        'short_model',
        'state',
        'age_category']
    return df.drop(columns_to_drop, axis=1)
 
 
models = (
    LogisticRegression(solver='liblinear'),
    RandomForestClassifier(),
    SVC()
)
 
 
imputer_transformer = Pipeline(steps=[
        ('imputer1', SimpleImputer(strategy='median'), make_column_selector(dtype_include=['int64', 'float64'])),
        ('imputer2', SimpleImputer(strategy='most_frequent'), make_column_selector(dtype_include=object))
    ])
 
encode_scale_transformer = Pipeline(steps=[
        ('scaler', StandardScaler(), make_column_selector(dtype_include=['int64', 'float64'])),
        ('encode', OneHotEncoder(handle_unknown='ignore'), make_column_selector(dtype_include=object))
    ])
 
 
X = df.drop(['price_category'], axis=1)
Y = df['price_category']
 
for model in models:
    pipe = Pipeline(steps=[
        ('filter', FunctionTransformer(filter_data)),
        ('outliers', FunctionTransformer(calculate_outliers)),
        ('imputer_transformer', imputer_transformer),
        ('predictors', FunctionTransformer(new_predictor)),
        ('encode_scale_transformer', encode_scale_transformer),
        ('filter_2', FunctionTransformer(filter_data2)),
        ('classifier', model)
    ])
 
    score = cross_val_score(pipe, X, Y, cv=4, scoring='accuracy')
    print(f'model:{type(model).__name__}, acc_mean: {score.mean():.4f}, acc_std:{score.std():.4f}')