02
Domain 02
Data & Leak-Proof Features
Leakage, pipelines, encoding, scaling, and missing data — the work that decides whether a model ships.
6 questions
- Q07Which feature must be removed before training, and why?The feature that makes your model brilliant offline and useless live.
- Q08What is the flaw, and how does scikit-learn help you avoid it?One-hot, ordinal, or target encoding for 3,000 zip codes?
- Q09What is the most appropriate way to encode zip_code?A column with too many categories — keep it, drop it, or encode it smarter?
- Q10Which feature is target leakage?Six candidate features for a readmission model. One of them is from the future.
- Q11What is wrong with Student A's approach?Two students, same scaler, different order of operations. One of them leaked.
- Q12What is the best way to handle this column in a pipeline?The blanks default at 31%. What do you do with a missing value that talks?