How long should be the sliding window for time series classification?
Gustavo Sánchez Hurtado
Award-Winning Engineer, Researcher & Educator | Digital Transformation: Control Systems, IoT, and Machine Learning | PLC/SCADA programmer | Python/MATLAB | Node Red | Global Speaker, Author & Podcaster
It is well-known that Sliding Window Length (SWL) directly affects classification performance. However, it is difficult to find reports of systematic studies about this topic, for different kinds of time series. One interesting example is presented in:
Just to have a broad idea, consider the problem showed in the figure above, with two classes: "slow" and "fast" sinusoidal.
Using a MLP classifier (Sklearn) it is possible to measure the following accuracy for increasing values of the SWL:
As you can see the accuracy increases with SWL, however also does the processing and detection time.
You can test the code here:
Feel free to leave your comments here below, I would be happy to answer.
At?MYWAI?we promote agile, explainable, reliable and affordable ML at the edge.