Dynamic Network Models for Forecasting

We have developed a probabilistic forecasting methodology through a synthesis of belief­ network models and classical time-series analysis. We present the dynamic network model (DNM) and describe methods for con­ structing, refining, and performing inference with this representation of temporal proba­ bilistic knowledge. The DNM representation extends static belief-network models to more general dynamic forecasting models by inte­ grating and iteratively refining contempora­ neous and time-lagged dependencies. We dis­ cuss key concepts in terms of a model for fore­ casting U.S. car sales in Japan.