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
458 stars 73 forks source link

Properly handle very large signals. Add Sampling #126

Closed antoinecarme closed 4 years ago

antoinecarme commented 4 years ago

PyAF should find a strategy to reduce the modeling time for very large signals.

Profiling shows that the for these signals, the most time consuming part is the AR modeling process (third part of the decompostion : scikit-learn , xgboost, keras etc).

It is possible to activate some kind of sampling for signals whose length is above a given threshold.

antoinecarme commented 4 years ago

Sampling is activated by default

lEngine.mOptions.mActivateSampling = True; # sampling can be used for very large time series
lEngine.mOptions.mSeed = True;
lEngine.mOptions.mSamplingThreshold = 80;  Time series larger than this threshold will be sampled.
antoinecarme commented 4 years ago

Closing