search

antoinecarme / pyaf

PyAF is an Open Source Python library for Automatic Time Series Forecasting built on top of popular pydata modules.
BSD 3-Clause "New" or "Revised" License
456 stars 73 forks source link

Lag generation process optimization #191

Closed antoinecarme closed 2 years ago

antoinecarme commented 2 years ago

PyAF generates a lot of lags for each cycle residue to compute additional signal components (AR, ARX, SVR, XGB, XGBX, ...)

These lags are generated on the same CPU for each cycle residue to compute a whole set of models.

The generated lags can be shared between all these models, using the same lags internal dataframe. Keras, XGBoost and Scikit-Learn models can use the same input numpy vectors.

This is a CPU time + memory optimization. No impact on forecast models and/or forecast values is expected.

Release date : 2022-07-14

antoinecarme commented 2 years ago

Ensure that PyAF is and remains Green #176

antoinecarme commented 2 years ago

This issue is linked with #190. Generating too many lags and merging the resulting dataframes leads to some re-indexing issues.

antoinecarme commented 2 years ago

FIXED.