【摘要】 采用遗传算法建立贝叶斯网络的优化学习结构,一直是贝叶斯网络研究倍受关注的课题。传统遗传算法的个体设计存在需要反复进行无环性检验的问题,降低了进化效率。针对这个问题,提出一种新的个体编码方式。考虑到进化过程中家族得分的可继承性,提出基于家族继承的结构评分改进算法,进而设计相应的改进遗传算法。实验结果表明,改进算法在BN建网精度与效率上都得到明显提升。更多还原

【Abstract】 Bayesian network structure learning based on genetic algorithm is always the subject of much attention.However,the individual design of the traditional genetic algorithm need to check the structure and ensure it acyclic repeatly,reducing the efficiency.For the problem,a new individual encoding is presented.In addition,taking into account inheritable family score,the improved algorithm based on family inheritance is proposed to mark structure,and then design the improved genetic algorithm is designed.Experimental results show that the improved algorithm makes better performances.更多还原