TANC-BIC结构学习算法的改进  

Improvement for TANC-BIC Structure Learning Algorithm

作　　者：程泽凯[1] 秦锋[1] 徐浩[1] 

机构地区：[1]安徽工业大学计算机学院,安徽马鞍山243002

出　　处：《计算机技术与发展》2006年第5期44-46,共3页Computer Technology and Development

基　　金：安徽省高等学校青年教师资助项目(2005JQ1079)

摘　　要：基于概率的贝叶斯分类器以其简单的结构和良好的性能受到重视,树扩展朴素贝叶斯分类器TANC应用较广。用TANC-BIC结构学习算法构建的分类器取得了成功,但TANC-BIC结构学习算法未考虑类节点的情况。文中提出了一种新的结构学习TANC-CBIC算法。并在贝叶斯分类器实验平台MBNC上编程实现。实验结果表明,改进算法分类准确率要高于由TANC-BIC和TANC-CMI结构学习算法构建的分类器,TANC-CBIC结构学习算法是有效的。Bayesian classifier based on probability theory has gained great attention ,because of its simple structure and good performance. TANC applies widely in practice. The classifier which was set up by the TANC- BIC structure- learning algorithm bad acquired success, but it didn＇t consider the class node. This paper suggests a new structure - learning algorithm called TANC - CBIC, makes experiment in MBNC experiment platform with programming TANC- CBIC algorithm. The results show that the accuracy of improver is better than algorithm based onTANC- BIC and TANC - CMI. The new structure - learning algorithm is effective.

关 键 词：树扩展朴素贝叶斯分类器 贝叶斯信息标准测度 结构学习 数据采掘 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]