I have generated some data with drift (abrupt, gradual, recurrent, and incremental). I want to compare some drift detectors algorithm on each but for that, I need the specific starting and ending position of where the drift occurs. Any idea?
For example, between 500 to 1000 and 2000 to 2500, I think the drifts occur but I'm not sure, and for what I want to do this needs to be precise. In red we can see concept probability, in blue the stream generated and other plots are drift detectors. Link to stream learn Documentation, github.
Looking at the concept_probabilities vectors try to identify how I can handle it but I wasn't... because sometimes it starts at 1, 0.99, 0.98 and fall into 0 but sometimes 0.01, and so on. Image of streaming data, concept probability, and drift detectors
