混合鱼群优化算法的贝叶斯网络结构学习

Structure Learning of Bayesian Network Via Hybrid Fish Swarm Optimization Algorithm

针对从数据集学习贝叶斯网络结构准确率不高的问题,提出了一种基于混合鱼群优化算法的结构学习算法。首先,利用互信息和最大似然树生成初始无向图;然后,由无向图的边随机生成初始种群,将粒子群算法的个体记忆和交流意识引入鱼群算法的行为模式,减小算法搜索行为的盲目性;最后,将优势遗传算法的变异和交叉算子应用于算法的寻优过程。仿真实验结果验证了改进后的算法具有更强的寻优能力。

In order to improve the accuracy of learning Bayesian network structure from the data set,a new method was proposed based on the hybrid Fish swarm optimization algorithm. Firstly,the initial undirected graph was generated by the mutual information and the maximum likelihood tree,which was the foundation of the initial population. Then the individual remembering capacity and communicating capacity of particle swarm optimization algorithm were introduced into the artificial fish swarm algorithm to avoid the blindness of searching. Finally,the algorithm referring to the mutation and crossover operator of adaptive genetic algorithm was used to improve the optimization process. Simulation experiment results show that the improved algorithm has better optimization ability.