Outlier detection routines (integrate PyOD)
I still think we should implement an outlier detection that is statistically more backed up. In slidingZscore we currently use a kind of scoring system approach, that flags a value if it was identified as an outlier in a sufficient number of rolling windows - whereby this minimum number and the size of that windows is parameter determined. That approach is quite intuitive but a little hard to determine and to calibrate in detail a think. Especially it doesnt account for propperly testing sets with multiple outliers in one test set.
There are statistical standart approaches available for testing for a arbitrary number of outliers that i would feel would be more solid to use. At least i think it would be proper to make available some basic tests from the Python Outlier detection package.