基于EM-GA改进贝叶斯网络的研究及应用  

Research on Bayesian network improved by EM-GA and its application

作　　者：金俊丽[1] 赵川[2] 杨洁[3] 

机构地区：[1]淮海工学院理学院,江苏连云港222005 [2]重庆大学机械工程学院,重庆400030 [3]重庆通信学院,重庆400035

出　　处：《计算机应用研究》2010年第4期1360-1362,共3页Application Research of Computers

基　　金：国家教育部"新世纪优秀人才支持计划"资助项目(NCET-07-0908)

摘　　要：为了解决软件风险分析中可能出现的数据不完整以及影响因素间关系复杂的问题,提出了一种改进贝叶斯网络的软件项目风险分析方法。将遗传算法和EM算法相结合得到EM-GA算法,利用EM-GA算法对软件项目分析过程中贝叶斯网络结构中的参数进行学习,同时优化网络结构,通过实例验证了该方法的有效性及可行性。In order to solve the problem of incomplete data and complex relations among influencing factors which may appear in the software risk analysis,this paper presented a software project risk analysis process based on Bayesian networks which has been improved.Firstly,presented a EM-GA algorithm based on genetic algorithm.Then,used the algorithm to optimize the Bayesian networks structures and solve Bayesian parameter learning.Finally,the experiment results show this algorithm provide a new method for software project risk analysis process.

关 键 词：贝叶斯网络 EM-GA算法 软件项目 风险分析 

分 类 号：TP311.52[自动化与计算机技术—计算机软件与理论]