Node Mutual Information Guided Improved PSO Bayesian Network Structure Learning Algorithm

Aiming at the problem that the Bayesian network structure learning algorithm has low spatial search efficiency and is prone to fall into the local optimum leading to the decline in the accuracy of the optimal network structure solution, the paper proposed a Bayesian structure learning algorithm based on improved particle swarm optimization (MIPSO-BN). The algorithm constructs the initial network structure through a mutual information guided optimization strategy to enhance the network search efficiency. Meanwhile, an improved particle swarm search strategy is used to enhance the diversity of the network search space and reduce the risk of the algorithm falling into a local optimum. Experiments show that the proposed method significantly outperforms the comparison algorithms in standard ASIA, CHILD, INSURANCE dataset tests, and the F1-Score is improved by an average of 12.96%, 47.95%, and 58.21%, respectively, when compared to the second-performing PC-PSO.