Time Series Machine Learning Quant Interview Guide
Time series machine learning quant interview guide for ordering, lag features, walk-forward splits, autocorrelation, leakage, drift, and examples.
Candidates modeling ordered financial, event, or macro data.
Time order is the constraint
Time-series machine learning must respect what was known at each point. Random splits, future features, and revised data can all leak information.
Lag features need clear timing
A lagged feature should be computed from data available before the prediction time. Publication delays and market close times may matter.
Concrete example
For a macro signal, a value released next week cannot be used to trade today even if the timestamp in a vendor table appears earlier.
Walk-forward validation is natural
Train on past data, validate on later data, then roll forward. This better reflects deployment than randomly mixing time periods.
Common mistakes
Candidates often build rolling features after splitting data. Feature engineering must be point-in-time aware before the validation result is trusted.
Practice the pattern
Use the LeetQuidity curriculum and calibration to turn this topic into a focused practice plan.