medium10 pts · first answer countsregressionfeature-engineering

Predicting a heavily skewed target

You are predicting customer lifetime spend, which is heavily right-skewed with a few huge values. A linear regression fits poorly and errors grow with the prediction size. What is a common first fix?

A.Model log(1 + spend) instead of raw spend, then transform predictions back carefullyB.Delete every customer with high spendC.Use accuracy as the evaluation metricD.Square the target to spread values out

Discussion 0

Join free to post in the discussion.

Predicting a heavily skewed target

Discussion 0

Related questions