question about the dynamic threshold
jlro001 opened this issue · comments
How did you come up with the calculating of the dynamic threshold?
Is it enlightened by other works?
Can you please kindly explain it?
@jlro001 thanks for the comment. You can find more information on how the dynamic threshold is set (briefly) in the video and (in detail) within the publication posted on the KDD website.
@jlro001 It was motivated by the fact that the residuals were non-Gaussian and a nonparametric approach allows for better generalization to other domains and datasets (with some tuning). Without too much overhead, all data points are scored in relation to the other data points, rather than relying on a distribution that doesn't fit the data to score each point. I haven't seen the approach in other works.
Firstly, thank u for responding.
Actually, my question is how you come up with the complex formula to determine the number of standard deviations above μ.
Are there some works related to this？
Thank u!
None that I'm aware of. I created it to try and balance the empirical issues I was seeing (as described in the paper).