基于改进遗传算法的贝叶斯网络结构学习OA

The scale of the Bayesian network structure search space increases exponentially with the number of nodes,which makes BN structure learning extremely difficult.Meanwhile,traditional genetic algorithm for BN structure learning easily falls into local optima.This paper adopts an improved genetic algorithm for Bayesian network structure learning.First,the population is initialized using conditional independence tests and mutual information.Then,the population is updated and iterated with the tournament selection strategy,improved crossover operator,and mutation operator in the search space generated by conditional independence tests.Finally,the optimal individual,i.e.,the final network structure,is selected according to the fitness function.Experimental results show that the proposed algorithm achieves higher accuracy and faster convergence speed in BN structure learning.