摘要
为克服K2算法在处理贝叶斯网络分类器(BayesianNetworkClassifier,BNC)结构学习中要求先指定适合节点次序的缺点,提出GA-K2算法,将基于选择性集成的整数编码遗传算法引入到K2算法中,使之能得到最佳节点次序并且网络结构收敛到全局最优.构建贝叶斯网络分类器进行分类,实验结果表明GA-K2算法优于随意指定节点顺序的K2算法.
To overcome the defect that K2 algorithm requires the suitable order of nodes in advance while dealing with the structure learning of Bayesian Network Classifier （ BNC ）, the algorithm GA-K2 is proposed which introduces the integer coding genetic algorithm based on selective ensemble concept to K2. It provides the guarantee of getting the best order of nodes and the convergence of Bayesian network structure for K2 in global optimization. The results of classification experiment by building BNC indicate that GA-K2 is better than K2 algorithm which is only with random order of nodes.
出处
《计算机辅助工程》
2006年第3期46-50,共5页
Computer Aided Engineering
基金
国家自然科学基金(60473115)
关键词
贝叶斯网络
分类器
结构学习
K2算法
遗传算法
Bayesian network
classifier
structure learning
K2 algorithm
genetic algorithm
