Lumpy demand forecasting on industrial demand data
Not many studies have been undertaken on forecasting lumpy demand. This exploratory study proposes two measures for characterizing lumpy demand. The lumpiness factor and the coefficient of skewness. These measures are applied to actual lumpy demand data of an electronics distributor operating in Monterrey, Mexico. Guidelines for short-term forecast performance are suggested. Three forecasting methods, Exponential smoothing, Croston's method, and Neural networks are applied to the data set. The performance of the forecasting is evaluated using three forecast performance evaluators namely Root Mean Squared Error, Mean Absolute Percentage Error and Mean Absolute deviation. The best forecasting technique to handle lumpy demand is presented based on the research conducted. ^
Bendore, Nithin Raj, "Lumpy demand forecasting on industrial demand data" (2004). ETD Collection for University of Texas, El Paso. AAIEP10523.