When you're dealing with timeseries in practice, you're usually working on figuring out the seasonal pattern. There are many ways to model this, but you may want to consider periodic features. They offer an elegant way to represent a repeating pattern and the SplineTransformer in scikit-learn gives you support for this directly. You even get some tuning knobs for gridsearch!
Video Chapters:
00:00 Introduction
01:21 Monthly Features
03:55 Improvements
06:20 Periodic Features
10:56 New Model
Docs for SplineTransformer:
scikit-learn.org/stable/modul...
The code for all of our videos can be found on this Github repository:
github.com/probabl-ai/youtube...
The code for this specific episode can be found here:
github.com/probabl-ai/youtube...
If you're keen to see more videos like this, you can follow us over at @probabl_ai.
Негізгі бет Generating Periodic Features for Seasonal Timeseries
Пікірлер: 6