ML Feature Importance Quant Interview Guide
ML feature importance quant interview guide for permutation importance, SHAP caveats, correlated features, stability, and examples.
Candidates interpreting tree models, linear models, or signal rankings.
Importance explains model usage
Feature importance describes how a model used inputs under a method. It does not automatically prove causality or economic truth.
Method choice changes the answer
Split gain, permutation importance, coefficients, and SHAP-style explanations answer different questions. Mention the method and its caveats.
Concrete example
If two features are highly correlated, a model may split importance between them or prefer one arbitrarily. Removing one can change the reported ranking.
Stability is evidence
Check whether importance is stable across time periods, folds, universes, and reasonable feature transformations. Unstable explanations need caution.
Common mistakes
Candidates often point to the top feature as the reason a strategy works. A stronger answer separates model diagnostics from causal claims.
Practice the pattern
Use the LeetQuidity curriculum and calibration to turn this topic into a focused practice plan.