Novel Bayesian Network Incremental Learning Method Based on Particle Swarm Optimization Algorithm

Abstract

Traditional bayesian network learning mainly uses batch learning method, that is, assuming that the training data can be acquired in one time, the whole learning process will end after learning these samples. But most real-world data is generated sequentially and incrementally over time. Therefore, it is necessary to improve batch learning into incremental learning to adapt to the reality. Many scholars have studied the incremental learning method [1, 2, 3] in the bayesian network field. In this paper, the discrete particle swarm optimization (PSO) algorithm is introduced into the structure learning of bayesian network, and a new incremental learning method based on PSO is proposed. Finally, the control variable method is used to compare the parameters of the algorithm, the number of iterations of a single learning and the number of particles, etc., to explore the influence of these factors on the incremental learning effect of bayesian network.