BlogIntermittent DemandLightGBM
Two-Stage Modeling for Intermittent Demand
Separating sale occurrence from sale magnitude can make sparse forecasting systems easier to reason about.
A two-stage demand model first estimates whether a sale will happen, then estimates the magnitude conditional on sale. This separation gives clearer diagnostics when most rows are zero.
It also lets the model optimize different objectives for occurrence and amount, which is often useful in product-level daily data.