hard20 pts · first answer countsinterpretabilityrandom-forest

Random ID tops feature importance

In a random forest, a meaningless high-cardinality user_id column ranks as the top feature by impurity-based importance. What should you conclude?

A.Impurity importance favours high-cardinality features; check permutation importanceB.The user_id truly causes the target and should stayC.Random forests cannot use numeric ID columnsD.Impurity importance is always reliable, so trust it

Discussion 0

Join free to post in the discussion.

Random ID tops feature importance

Discussion 0

Related questions