VARIABLE AGGREGATION METHOD BY FORMING KEY NODES IN DESIGNING BAYESIAN NETWORKS

Abstract

In the current conditions of increasing complexity of information systems and data volumes, methods for building effective models for uncertainty analysis and decision-making are of particular importance. Bayesian networks are one of the most common tools in this area, since they allow combining probabilistic dependencies between variables with a logical structure of cause-and-effect relationships. However, with an increase in the number of variables and relationships, the problem of excessive complexity of the model arises, which complicates both its construction and further analysis. One of the promising approaches to solving this problem is the aggregation of variables by forming key nodes that represent the most significant characteristics of the system. This method allows to reduce the network’s dimensionality, while maintaining its informativeness and adequacy of the studied processes reflection. As a result, a more compact and computationally convenient structure is created, which contributes to increasing the efficiency of both the construction and Bayesian networks use in practical problems. This paper discusses the variable aggregation method theoretical foundations, the key node formation principles, and the features of applying this approach in designing Bayesian networks.