Forecasting demand for slow-moving items in case of reporting errors
Abstract
The paper considers the problem of demand forecasting for slow-moving items in case of reporting errors. A generalization of the beta-binomial demand model is proposed that takes into account possible distortions in the learning sample. Properties of the underlying probability distribution are derived. For this new model, algorithms that provide consistent estimators of the model parameters as well as mean square error optimal forecasts when used for historical demand data with reporting errors are developed. An example for slow-moving car parts is given to illustrate the proposed demand forecasting approach.