Dynamic Network Models for Forecasting

Paul Dagum, Adam Galper, and Eric Horvitz

Click here to access pdf file.


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 constructing, refining, and performing inference with this represetnation of temporal probabilistic knowledge. The DNM representation extends static belief-network models to more general dynamic forecasting models by integrating and iteratively refining contemporaneous and time-lagged dependencies. We discuss key concepts in terms of a model for forecasting U.S. car sales in Japan.

Keywords: dynamic network models, dynamic Bayesian networks, temporal reasoning.

In: Proceedings of the Eighth Annual Conference on Uncertainty in Artificial Intelligence (UAI 1992), July 17-19, 1992, Stanford University, Stanford, CA 1992.