A stochastic state space model for prediction of product demand

Abstract

This paper is concerned with the development of a fixed price, supply/demand market model which can be used to predict demand for customer premises telephone equipment. A state space approach is used to model system dynamics and a Kalman filter is used for estimation. The model is nonlinear, and provides for nonstationary statistical characterization of the elements. The formulation indicates theoretically that, given perfect input (driving force) data, predictions could be highly inaccurate using linear models (even if they are dynamic) or nonlinear models which assume stationary statistics. The conceptual framework afforded by state space provides a vehicle for structuring more accurate models to predict product demand than do conventional approaches. Finally, the general model is suitable for predicting product demand in a wide range of markets.