Bayesian Forecasting of Spare Parts Using Simulation

Abstract

Forecasting demand plays a major role in many service and manufacturing organizations. Forecasts help in the scheduling of taskforce, obtaining higher service levels for the customer, and determining resource requirements among many others (Makridakis et al. 1998). Forecasting accuracy is an increasingly important objective in most firms and, in particular, plays a key role in forecasting lumpy demand. According to many authors (e.g., Wacker and Sprague 1998; Zotteri and Kalchsdmidt 2007a), the accuracy of forecasts depends sensitively on the quantitative technique used, thus, this chapter has been motivated by an increasing need for applying and formulating new tools for demand forecasting, and in particular for the case of lumpy demand. As suggested by the work of Caniato et al. (2005) and Kalchsdmidt et al. (2006), it is necessary to propose forecasting techniques that not only take into account the time series, but also the structure of the demand-generating process (non-systematic variability). For this reason, in this chapter we illustrate how to apply simulation techniques and Bayesian statistics in a model that takes into account particular characteristics of the system under study.