Core ML Concepts · intermediate · concept 32 of 311 · Updated Nov 4, 2026

🧠Data Leakage

When information from outside the training set sneaks into training, so the model aces validation and faceplants in production. Classic forms: preprocessing before the split, target information hiding in a feature, duplicate rows across splits, and peeking into the future on time series. The number-one practical reason ML projects report great numbers and then fail.