Data Leakage

91 min
Block 7 — Train, Validation and Test
Objective
define data leakage rigorously; separate its two disjoint branches — target leakage, where a feature encodes information about the label that will not exist when the prediction has to be made, and protocol leakage, where information flows from the evaluation data into training through the procedure itself; recognize the eighteen concrete forms these two branches take; measure, on executed code, the size of the illusion each one produces and the honest figure it replaces; run a systematic audit; and build a learning architecture that eliminates whole classes of leakage mechanically rather than by vigilance.
Estimated duration
90 minutes
Prerequisites
chapters 019 (imputation), 022 (categorical encoding), 023 (feature scaling), 024 (feature engineering), 025 (feature selection), 026 (train/validation/test) and 027 (stratification, groups, time)
Associated quizzes
028.1-quiz-definition-and-severity.md to 028.7-quiz-structural-protection.md

Preview — the rest of the lesson is for enrolled readers.

Already enrolled with a code?

Your access is tied to your account, not to this link. Sign in with the same email you used in class: your course is waiting, no need to enter the code again.

Sign inNo account yet? Create one
This lesson is part of the “Splitting the Data: Train, Validation, Test” module

The first modules of the course are open to everyone. For the rest you have three options: buy this course once and for all, subscribe, or enter the code handed out in class.

Are you a student on this course?

The code is tied to your account: sign in or create an account and it will be applied automatically when you come back.

No account yet? Create one