What is data leakage in machine learning?
Interviewer Expectations for this Question
01
Core Competency
Assesses fundamental understanding of Machine Learning conventions, runtime behavior, and memory/performance considerations.
02
Evaluation Criteria
Hiring managers look for precision, avoidance of ambiguous jargon, and ability to explain trade-offs under real production conditions.
Comprehensive Model Answer
Verified Solution
Data leakage is when information unavailable at prediction time seeps into training, producing optimistic validation scores that collapse in production.
Common sources:
- Target leakage: a feature that is a proxy for the label, such as account_closed_date when predicting churn.
- Train-test contamination: fitting scalers, imputers, target encoders or SMOTE on the full dataset before splitting.
- Temporal leakage: using future data to predict the past, as with random splitting of time series.
- Group leakage: the same entity appears in both train and test.
Prevention:
- Split first, then fit all preprocessing on training folds only, ideally inside a pipeline.
- Use time-based splits for temporal problems.
- Audit features for whether they are known at prediction time.
- Use grouped splits when rows share an entity.
Pipeline([("scaler", StandardScaler()), ("clf", LogisticRegression())])
Leakage often shows as suspiciously high accuracy, so treat such results with suspicion.
Candidate Response Strategy & Interview Tips
- Start with a concise one-sentence summary: Deliver a direct, confident answer first before expanding into nuances.
- Demonstrate real-world trade-offs: Discuss where this approach excels and when you would avoid it in production systems.
- Discuss complexity & edge cases: Proactively explain time/space complexity or boundary conditions (null values, scale limits).
- Prepare for interviewer follow-ups: Technical hiring panels frequently probe deeper into concurrency, backward compatibility, or alternative libraries.