juniorTime Series
What is a lag feature in time series modeling?
Updated May 15, 2026
Short answer
A lag feature uses past values of a time series as inputs for predicting future values.
Deep explanation
Lag features transform time series into supervised learning format by shifting data backward. For example, lag-1 uses previous timestep value as input. This helps machine learning models capture temporal dependencies explicitly.
Real-world example
Predicting tomorrow’s sales using yesterday’s sales.
Common mistakes
- Using future data in lag creation (data leakage).
Follow-up questions
- What is data leakage in time series?
- How many lag features should be used?