Last active
February 27, 2019 14:32
-
-
Save eliorc/e847b0a8641d8613bcb43f2fe6daa33c to your computer and use it in GitHub Desktop.
Generate cyclic time features on pandas.DataFrame using a pandas.DateTime column
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
from itertools import count | |
import pandas as pd | |
import numpy as np | |
def cyclical_encoder(data: pd.DataFrame, | |
time_column: str, | |
time_unit: str, | |
normalize_val: float, | |
label_suffix: str) -> pd.DataFrame: | |
""" | |
Encodes a cyclical time unit using sinus and cosine | |
""" | |
# Get time values | |
time_values = getattr(data[time_column].dt, time_unit) | |
sin_time = np.sin(2*np.pi*time_values / normalize_val) | |
cos_time = np.cos(2*np.pi*time_values / normalize_val) | |
# Determine labels, check if already exists | |
sin_label = 'sin_{}_'.format(time_unit) | |
cos_label = 'cos_{}_'.format(time_unit) | |
if label_suffix: | |
sin_label += str(label_suffix) | |
cos_label += str(label_suffix) | |
else: | |
for n in count(): | |
if not sin_label + str(n) in data.columns: | |
break | |
sin_label += str(n) | |
cos_label += str(n) | |
return pd.concat([data, pd.DataFrame(columns=[cos_label, sin_label], data=np.array([cos_time, sin_time]).T)], axis=1) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment