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

PyAF 5.0 Final Touch 1 : discard some non-significant components #230

Closed antoinecarme closed 1 year ago

antoinecarme commented 1 year ago

Some cycles/ARs are almost zero. Detect this earlier and decrease the number of candidates accordingly. This is a small optimization.

Can discard all cycles if the trend residue is small (max - min < 0.001).

Can discard all ARs if the cycle residue is small (max - min < 0.001).

the criterion (max - min < 0.001) has to be kept as simple as possible for robustness. Do not take into account outliers etc (this is a different task and needs some real work).

The threshold 0.001 is reasonable as the signal is re-scaled (to be in [0, 1]) based on min/max before each transformation.

antoinecarme commented 1 year ago

PyAF 6 :

Outliers in Time Series : https://github.com/antoinecarme/pyaf/blob/master/TODO.md