SCM供应链 8-Forecasting-07.ppt

Prepared by WANG Jun qwqw  Introduction Planning and controlling logistics/supply chain activities require accurate estimates of the product and service volumes. These estimates are typically in the form of forecasts and predictions. This chapter overviews those techniques most likely to be directly used for logistics planning and control, and discuss demand forecasting. When the uncertainty of the predictive variable is so high that standard forecasting techniques lead to unsatisfactory results, other planning approaches, collaborate forecasting (prediction), are needed. Nature of Forecasting Spatial vs. temporal demand Logistics has both space and time dimensions. Forecasting techniques should reflect geographic differences that may affect demand patterns. Forecasting techniques may differ depending on whether all demand is forecasted and then disaggregated by geographic (top-down forecasting) or whether each geographic location is forecasted separately and aggregated (bottom-top forecasting) Demand variation is a result of growth or decline in sales rates, seasonality in the demand pattern. Lumpy vs. regular demand When demand is “regular”, it will typically be represented by one of the general patterns. Demand patterns can usually be decomposed into trend, seasonal and random components. When random variations are a small portion of the remaining variation in the time series, the popular forecasting methods can get good results. When demand is intermittent, low volume overall and a high degree uncertainty, the time series is said to be lumpy (irregular). Lumpy demand patterns are particularly difficult to forecast using the more popular techniques. Derived vs. independent demand Independent demand is generated from many customers, most of which individually purchase only a small fraction of the total volume. Derived demand is derived from the requirements specified in a production schedule. Statistical forecasti