Automated Time Series Feature Engineering
PyChronoBoost is a utility Python package designed for time series data analysis and forecasting. It offers robust functionality for imputing missing values, generating and selecting features, and handling various complexities associated with time series data.
- Time Series Imputation: Efficiently handles missing data in time series, both in terms of values and time steps.
- Feature Generation: Automatic generation of relevant features from time series data.
- Feature Selection: Utilizes algorithms like XGBoost to select the most significant features for your analysis.
Install PyChronoBoost using pip:
pip install git+https://github.com/jimmyyih518/PyChronoBoostHere's a quick example to get you started with PyChronoBoost:
import pandas as pd
from pychronoboost.timeseries import TimeSeriesData
# Sample time series data
data = pd.DataFrame({
'time': pd.date_range(start='2021-01-01', periods=5, freq='D'),
'value': [1, None, 3, 4, 5]
})
# Initialize TimeSeriesData object
ts_data = TimeSeriesData(data, timestep_column='time')
# Process features
processed_data = ts_data.process_timeseries_features(
feature_columns=['value'],
target_column='value',
value_impute_strategy='last',
max_window_size=3,
feature_selector_model='XGB',
max_features=5
)
print(processed_data)
For more detailed information and examples, please refer to the example notebook.
Contributions to PyChronoBoost are welcome! Please contact the author for more information.
PyChronoBoost is released under the Apache License.
For any queries or suggestions, feel free to open an issue on GitHub or contact me directly at [jimmyyih@ualberta.com].
I hope PyChronoBoost makes your time series data analysis more efficient and insightful!