Skip to content
ML Academy

Introduced on day 12

Out-of-bag (OOB) estimate

An error estimate for a bagged ensemble in which each row is predicted only by the trees whose bootstrap sample did not contain it.

In detail

OOB scoring gives a validation-like error without a separate hold-out set, which is convenient for forests. It assumes rows are exchangeable, so it is optimistic for time-ordered or grouped data where neighbouring rows share information with the trees that did include them. Use a time-ordered or group-aware split to confirm (Day 15).