A data science team is building a demand forecasting model using a dataset with daily sales records. They want to capture weekly seasonality and recent trend as input features. Which feature engineering techniques are most appropriate?

Question

Accepted Answer

B. Lag features such as sales_lag_7 and rolling window statistics such as rolling_mean_7. For time-series forecasting, lag features capture the value from N periods ago (capturing autocorrelation), and rolling window statistics such as a 7-day rolling mean capture recent trend and seasonality — both are standard feature engineering techniques for tabular time-series models.

A data science team is building a demand forecasting model using a dataset with daily sales records. They want to capture weekly seasonality and recent trend as input features. Which feature engineering techniques are most appropriate?

Related Questions

Discussion