A graph partitioning approach for Bayesian Network structure learning

Abstract

Structure learning of Bayesian Network is one of important topics in machine learning and widely applied in expert system. The traditional algorithms for structure learning are usually focused on the entire nodes in BN. It is difficult to learn the structure efficiently from the huge amounts of data. In reality, BN as a special inference network and the community also exists in BN. To achieve this goal, we propose Graph Partitioning Approach for BN Structure Learning. Firstly, we get the skeleton of BN by conditional dependence test. Secondly, skeleton is divided into some communities. Thirdly, the structure of every community is learned and the edges between communities are determined by BIC (Bayesian Information Criterion) score function. Numerical experiments on the standard network show that our proposed algorithm can greatly reduce the time cost of structure learning and have more accuracy.