Lumpy demand characterization and forecasting performance using self-adaptive forecasting models and Kalman Filter

The purpose of the study is to propose a systematic approach for lumpy demand characterization of historical and projected demand patterns to determine the extent of demand variability. Two proposed techniques are presented to improve forecast performance of lumpy demand observations, self-adaptive forecasting model and Kalman Filter. These techniques are described and applied on industrial demand data. A discussion of model building procedure of these modeling approaches is presented. The results indicate that these approaches exhibit a substantial forecasting performance improvement over traditional lumpy forecasting techniques