MAS 2012 Proceeding

Intermittent demand forecasting and stock control: an empirical study

Authors:   Adriano O. Solis, Letizia Nicoletti, Somnath Mukhopadhyay, Laura Agosteo, Antonio Delfino, Mirko Sartiano

Abstract

Statistical accuracy measures are generally used to assess the effectiveness of demand forecasting methods. In the final analysis, however, these methods should be judged according to whether they actually lead to better inventory control performance. We empirically evaluate four methods (simple moving average, single exponential smoothing, Croston?s method, and the Syntetos-Boylan approximation) in terms of statistical forecast accuracy and, more importantly, inventory system efficiency. We apply four forecasting methods to an industrial dataset involving more than 1000 stock- keeping units of a firm in the professional electronics sector. Demand is often intermittent, erratic, or both (i.e., lumpy). We devise and use a two-stage distribution involving uniform and negative binomial distributions to model the actual demand distribution, where possible. We then simulate the stock control performance of a (T,S) inventory system with respect to target customer service levels.

I3M  Scientific Sponsors

I3M  Industrial Sponsors

I3M  Media Sponsors