Date conversion and feature engineering for NYC cab data

nyc_date_conversion.py

import pandas as pd 
from datetime import datetime
import numpy as np 
import re

# To datetime conversion 
def to_datetime(x: str) -> datetime:
    """
    Converts a string to a datetime object
    An example of the string is 2010-02-02 17:24:55
    """
    # Inspecting whether x is datetime 
    if isinstance(x, datetime):
        return x
        
    try:
        # Dropping the UTC part from the date strings
        x = re.sub(' UTC', '', x)
        return datetime.strptime(x, '%Y-%m-%d %H:%M:%S')
    except:
        print(f"Error converting {x} to datetime")
        return pd.to_datetime(x)

def create_date_vars(
    d: pd.DataFrame, 
    date_var: str = 'pickup_datetime'
    ) -> pd.DataFrame:
    """
    Creates the datetime variables

    Creates the following columns 
        * pickup_dayofweek - The day of the week at pickup time
        * pickup_hour - The hour of the day at pickup time
        * pickup_dayofyear - The day of the year at pickup time
        * pickup_hour_sin, pickup_hour_cos - The sine and cosine of the hour of the day
        * pickup_dayofyear_sin, pickup_dayofyear_cos - The sine and cosine of the day of the year
    """
    # Infering the day of the week from pickup_datetime
    d[date_var] = [to_datetime(x) for x in d[date_var]]
    d['pickup_dayofweek'] = d[date_var].dt.dayofweek

    # Infering the hour of the day from pickup_datetime
    d['pickup_hour'] = d[date_var].dt.hour

    # Creating a new variable for the day of the year
    d['pickup_dayofyear'] = d[date_var].dt.dayofyear

    # Ensuring a monotonic relationship between pickup_hour and pickup_dayofyear
    d['pickup_hour_sin'] = np.sin(2 * np.pi * d['pickup_hour']/23.0)
    d['pickup_hour_cos'] = np.cos(2 * np.pi * d['pickup_hour']/23.0)

    d['pickup_dayofyear_sin'] = np.sin(2 * np.pi * d['pickup_dayofyear']/365.0)
    d['pickup_dayofyear_cos'] = np.cos(2 * np.pi * d['pickup_dayofyear']/365.0)

    return d