I am developing a use case to detect outliers on logons for a specific app using Smart Outlier Detection Assistant in MLTK app.
There is the Outlier Tolerance Threshold parameter in the Learn stage which I am unsure how to use.
The doc states "Adjust as needed based on the number of expected outliers."
how can someone know how many outliers they are expected ? To me, it doesn't makes sense as to how or why someone would already know this. Its what the usecase to indicate the number of outliers or maybe I am misinterpreting something.
Can someone please explain or direct me to some good documentation which properly explains this ?