expand_date_column.py

def transform_timestamp_to_features(timestamp):
    # Convert timestamp to datetime
    dt = pd.to_datetime(timestamp)

    # Initialize a dictionary to store features
    features = {}

    # Date Features
    features["year"] = dt.year
    features["quarter"] = dt.quarter
    features["month"] = dt.month
    features["week"] = dt.isocalendar().week
    features["day"] = dt.day
    features["doy"] = dt.dayofyear
    features["weekday"] = dt.weekday()
    features["is_business_day"] = int(dt.weekday() < 5)
    features["dow"] = dt.dayofweek
    features["wom"] = (dt.day - 1) // 7 + 1
    features["days_in_month"] = pd.Period(year=dt.year, month=dt.month, freq='M').days_in_month
    features["season"] = (dt.month % 12 // 3) + 1  # 1: Winter, 2: Spring, 3: Summer, 4: Fall
    features["is_leap_year"] = int(dt.is_leap_year)

    # Time Features
    features["hour"] = dt.hour
    features["minute"] = dt.minute
    features["second"] = dt.second

    # Binary flags for specific times and dates
    features["is_morning"] = int(6 <= dt.hour < 12)
    features["is_afternoon"] = int(12 <= dt.hour < 18)
    features["is_evening"] = int(18 <= dt.hour < 24)
    features["is_night"] = int(0 <= dt.hour < 6)
    features["is_lunch_time"] = int(dt.hour == 12 and dt.minute < 30)
    features["is_midnight"] = int(dt.hour == 0 and dt.minute == 0)
    features["is_early_morning"] = int(5 <= dt.hour < 9)
    features["is_late_night"] = int(21 <= dt.hour < 24)
    features["is_midday"] = int(dt.hour == 12 and dt.minute == 0)
    features["is_morning_commute"] = int(dt.hour == 8 and dt.minute <= 30)
    features["is_evening_commute"] = int(dt.hour == 17 and dt.minute >= 30)
    features["is_first_hour_of_day"] = int(dt.hour == 0)
    features["is_last_hour_of_day"] = int(dt.hour == 23)

    # Periods of the week
    features["is_weekend"] = int(dt.weekday() >= 5)
    features["is_weekday_morning_rush"] = int(dt.weekday() < 5 and dt.hour == 8)
    features["is_weekday_evening_rush"] = int(dt.weekday() < 5 and dt.hour == 17)
    features["is_weekday_start"] = int(dt.weekday() < 5 and dt.hour == 9 and dt.minute == 0)
    features["is_weekday_end"] = int(dt.weekday() < 5 and dt.hour == 17 and dt.minute == 0)
    features["is_weekend_start"] = int(dt.weekday() == 4 and dt.hour == 18)
    features["is_weekend_end"] = int(dt.weekday() == 6 and dt.hour == 23 and dt.minute == 59)
    features["is_week_start"] = int(dt.weekday() == 0)
    features["is_week_end"] = int(dt.weekday() == 4)

    # Periods of the year
    features["is_month_start"] = int(dt.is_month_start)
    features["is_month_end"] = int(dt.is_month_end)
    features["is_quarter_start"] = int(dt.is_quarter_start)
    features["is_quarter_end"] = int(dt.is_quarter_end)
    features["is_year_start"] = int(dt.is_year_start)
    features["is_year_end"] = int(dt.is_year_end)

    # Work and non-work hours
    features["is_work_hour"] = int(dt.weekday() < 5 and 9 <= dt.hour < 17)
    features["is_non_work_hour"] = int(not (dt.weekday() < 5 and 9 <= dt.hour < 17))

    # Periods of the hour
    features["is_first_minute_of_hour"] = int(dt.minute == 0)
    features["is_last_minute_of_hour"] = int(dt.minute == 59)
    features["is_half_hour"] = int(dt.minute == 30)
    features["is_quarter_past"] = int(dt.minute == 15)
    features["is_quarter_to"] = int(dt.minute == 45)

    # Aggregated and cyclic features
    features["minute_of_day"] = dt.hour * 60 + dt.minute
    features["second_of_day"] = features["minute_of_day"] * 60 + dt.second
    features["minute_sin"] = np.sin(2 * np.pi * dt.minute / 60)
    features["minute_cos"] = np.cos(2 * np.pi * dt.minute / 60)
    features["hour_sin"] = np.sin(2 * np.pi * dt.hour / 24)
    features["hour_cos"] = np.cos(2 * np.pi * dt.hour / 24)
    features["week_of_month"] = (dt.day - 1) // 7 + 1
    features["dayofweek_sin"] = np.sin(2 * np.pi * dt.weekday() / 7)
    features["dayofweek_cos"] = np.cos(2 * np.pi * dt.weekday() / 7)

    # Additional time features
    features["minute_of_half_hour"] = dt.minute % 30
    features["minute_of_quarter"] = dt.minute % 15
    features["second_of_hour"] = dt.second + dt.minute * 60
    features["is_am"] = int(dt.hour < 12)
    features["is_pm"] = int(dt.hour >= 12)
    features["is_early_morning"] = int(5 <= dt.hour < 9)
    features["is_late_night"] = int(21 <= dt.hour < 24)
    features["is_midday"] = int(dt.hour == 12 and dt.minute == 0)
    features["is_late_afternoon"] = int(15 <= dt.hour < 18)
    features["is_evening_peak"] = int(18 <= dt.hour < 20)
    features["is_early_evening"] = int(18 <= dt.hour < 21)

    return features
  
  
  def expand_date_columns(df, column_name):
    features_df = df[column_name].apply(transform_timestamp_to_features).apply(pd.Series)
    df = pd.concat([df, features_df], axis=1)
    return df