A framework for the analysis of dynamic processes based on Bayesian networks and case-based reasoning

Bayesian networks are knowledge representation schemes that can capture pro babilistic relationships among variables and perform probabilistic inferenc e. Arrival of new evidence propagates through the network until all variabl es are updated. At the end of propagation, the network becomes a static sna pshot representing the state of the domain for that particular time. This w eakness in capturing temporal semantics has limited the use of Bayesian net works to domains in which time dependency is not a critical factor. This pa per describes a framework that combines Bayesian networks and case-based re asoning to create a knowledge representation scheme capable of dealing with time-varying processes. Static Bayesian network topologies are learned fro m previously available raw data and from sets of constraints describing sig nificant events. These constraints are defined as sets of variables assumin g significant values. As new data are gathered, dynamic changes to the topo logy of a Bayesian network are assimilated using techniques that combine si ngle-value decomposition and minimum distance length. The new topologies ar e capable of forecasting the occurrences of significant events given specif ic conditions and monitoring changes over time. Since environment problems are good examples of temporal variations, the problem of forecasting ozone levels in Mexico City was used to test this framework. (C) 1998 Elsevier Sc ience Ltd. All rights reserved.