Finding Plateaus in TS data

[Mopele]

Hey, I want to find plateaus in my time series data. As I understand stumpy, it is a tool to find patterns and anomalies. But can I tell it somehow to find plateaus? Is there a Example for that?

Thanks!

[seanlaw]

Hey, I want to find plateaus in my time series data. As I understand stumpy, it is a tool to find patterns and anomalies. But can I tell it somehow to find plateaus? Is there a Example for that?

Thanks!

@Mopele Thank you for the question. Since the term plateau may be somewhat overloaded and can be interpreted in multiple ways, would you mind providing a visual example of what you mean by a "plateau"?

[Mopele]

Yes, I will provide visuals. Currently I am detecting "plateaus" by looking at the derivate of the values and if the derivate doesnt spike I asume it is a plateau.

I have a easy dataset and a hard one. In the hard one you cant see the plateaus good but they look similar to the one in the easy example.

Easy

Hard:

[seanlaw]

So, given the easy and hard time series above, if we use these as separate inputs to some analysis, would would be the expected output(s)? Would it be a simple threshold value like 200 for the easy case and 5000 for the hard case?

My initial thought is that STUMPY would not be much help here but I would need to better understand your goal(s).

[Mopele]

So right now I have some code that takes the timeseries as input and outputs a list of tupels with a startindex and endindex of a plateau.

Would such a result be possible with stumpy?

It would also help if I could somehow find the amplitude of the signals just for the flat parts. This would also get me better results.

I thought that with stumpy I could give it a flat section and it could output me similar sections so i can find the amplitude of the sections, but i am still new to stumpy...

[seanlaw]

I might be overlooking something obvious but, unfortunately, I don't think this is a great use case for STUMPY.

[Mopele]

Ok, thanks! So I guess then stumpy will not help me out so much