easy5 pts · first answer countsml-pipelinesdata-splitting

Splitting a rare-event dataset

Fraud is 0.5% of rows. A plain random split leaves the test set with very few frauds and noisy metrics. What helps?

A.Removing most fraud rows from the test setB.A stratified split that keeps the fraud rate in each partC.Using only 1% of the data for trainingD.Splitting by the alphabetical order of names

Discussion 0

Join free to post in the discussion.

Splitting a rare-event dataset

Discussion 0

Related questions