I have a similar use case requiring to compare two distributions, or same column from 2 date periods to detect significant change (i.e. anomaly detection).
I understand a well known method KL Divergence can be used to compare 2 distributions.
But seems that requires the 2 distributions to have same set of bins/buckets.
Can KLL be configured to produce such type of output that can be used for KL-Divergence?
Or is there other methods (distance function) for comparing 2 distributions and produce a consistent number indicating between them?
sscdotopen
commented
3 years ago
I have a similar use case requiring to compare two distributions, or same column from 2 date periods to detect significant change (i.e. anomaly detection). I understand a well known method KL Divergence can be used to compare 2 distributions. But seems that requires the 2 distributions to have same set of bins/buckets.
Can KLL be configured to produce such type of output that can be used for KL-Divergence? Or is there other methods (distance function) for comparing 2 distributions and produce a consistent number indicating between them?
Thanks in advance