Int. J. of Industrial and Systems Engineering   »   2012 Vol.11, No.1/2

 

 

Title: A Bayesian approach to forecast intermittent demand for seasonal products

 

Authors: Mohammad Anwar Rahman; Bhaba R. Sarker

 

Addresses:
Department of Industrial Engineering Technology, The University of Southern Mississippi, 118 College Drive, Hattiesburg, MS 39406, USA
Department of Industrial Engineering, Louisiana State University, Baton Rouge, LA 70803, USA

 

Abstract: This paper investigates the forecasting of a large fluctuating seasonal demand prior to peak sale season using a practical time series, collected from the US Census Bureau. Due to the extreme natural events (e.g. excessive snow fall and calamities), sales may not occur, inventory may not replenish and demand may set off unrecorded during the peak sale season. This characterises a seasonal time series to an intermittent category. A seasonal autoregressive integrated moving average (SARIMA), a multiplicative exponential smoothing (M-ES) and an effective modelling approach using Bayesian computational process are analysed in the context of seasonal and intermittent forecast. Several forecast error indicators and a cost factor are used to compare the models. In cost factor analysis, cost is measured optimally using dynamic programming model under periodic review policy. Experimental results demonstrate that Bayesian model performance is much superior to SARIMA and M-ES models, and efficient to forecast seasonal and intermittent demand.

 

Keywords: intermittent time series; SARIMA; seasonal autoregressive integrated moving average; peak demand; dynamic programming; Bayesian model; intermittent demand; seasonal products; forecasting.

 

DOI: 10.1504/IJISE.2012.046660

 

Int. J. of Industrial and Systems Engineering, 2012 Vol.11, No.1/2, pp.137 - 153

 

Available online: 17 Apr 2012

 

 

Editors Full text accessAccess for SubscribersPurchase this articleComment on this article