src/custom_transformers.py

import pandas as pd
import numpy as np
from sklearn.base import BaseEstimator, TransformerMixin



class FeatureExtractor(BaseEstimator, TransformerMixin):
    def fit(self, X, y=None):
        return self

    def transform(self, X):
        X_ = X.copy()
        X_["date"] = pd.to_datetime(X_["date"])
        X_["day_of_week"] = X_["date"].dt.dayofweek
        X_["month"] = X_["date"].dt.month
        X_["day_of_year"] = X_["date"].dt.dayofyear
        return X_[["name", "price", "rating", "smart_score", "review_count", "day_of_week", "month", "day_of_year"]]

def add_cyclical(df, col, max_val):
    radians = 2 * np.pi * df[col] / max_val
    return pd.DataFrame({
        f"{col}_sin": np.sin(radians),
        f"{col}_cos": np.cos(radians)
    })

def cyclical_features(X):
    X_ = X.copy()
    dow = add_cyclical(X_, "day_of_week", 7)
    month = add_cyclical(X_, "month", 12)
    doy = add_cyclical(X_, "day_of_year", 365)
    X_ = X_.drop(columns=["day_of_week", "month", "day_of_year"])
    return pd.concat([X_, dow, month, doy], axis=1)