hard20 pts · first answer countsmodel-monitoringdata-drift

Drift on 19,000 pincodes

You want a drift check on "pincode", which has about 19,000 values, many with very few rows. What works well?

A.Run a chi-squared test over all 19,000 raw valuesB.Treat pincode as a number and run a KS test on itC.Track top-k values plus "other", or roll up to districtD.Skip drift checks for every categorical feature

Discussion 0

Join free to post in the discussion.

Drift on 19,000 pincodes

Discussion 0

Related questions