Instruction: Describe the process and considerations for feature engineering when working with time-series data.
Context: This question evaluates the candidate's ability to handle specialized data types and their skill in enhancing model performance through feature engineering.
Official answer available
Preview the opening of the answer, then unlock the full walkthrough.
The way I'd approach it in an interview is this: For time-series data, I usually build features that respect order and causality. That means lag features, rolling statistics, seasonality markers, calendar features, trend indicators, and domain-specific event signals that would have been known at prediction time. I am...