A MODIFIED BOOTSTRAP METHOD FOR INTERMITTENT DEMAND FORECASTING FOR SPARE PARTS

A critical area in inventory management is the forecasting of parts or items. Intermittent demand forecasting (IDF) is a highly important forecasting method in inventory management, though it is difficult to predict intermittent demand because of its characteristics. There are two approaches to IDF: a parametric approach that estimates demand like traditional forecasting methods, and a non-parametric approach that estimates distribution of demands via resampling technique. Bootstrap method is considered to be one of the key non-parametric methods available today. Despite its usefulness, however, application of conventional bootstrap methods in IDF do not take into consideration any existing dependent structures in lead time demand, leading to inaccurate forecasts. In this paper, we highlight the unique characteristics of IDF and suggest a new bootstrap method that takes these characteristics into consideration. We conclude by demonstrating the applicability of this new method through a simple case experiment and its results