Data leakage occurs when information unavailable at prediction time, including information from nominally held-out observations, influences model fitting or selection. It makes estimated out-of-sample performance optimistically biased.
Articles by others on the same topic
There are currently no matching articles.