计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (16): 34-36.

• 理论研究 • 上一篇    下一篇

基于领域知识的贝叶斯网络结构学习算法

莫富强,王 浩,姚宏亮,俞 奎   

  1. 合肥工业大学 计算机与信息学院,合肥 230009
  • 收稿日期:2007-09-28 修回日期:2008-03-14 出版日期:2008-06-01 发布日期:2008-06-01
  • 通讯作者: 莫富强

Domain knowledge based Bayesian networks structure learning algorithm

MO Fu-qiang,WANG Hao,YAO Hong-liang,YU Kui   

  1. School of Computer and Information,Hefei University of Technology,Hefei 230009,China
  • Received:2007-09-28 Revised:2008-03-14 Online:2008-06-01 Published:2008-06-01
  • Contact: MO Fu-qiang

摘要: 针对SEM算法在缺省数据学习中存在精度偏低和收敛速度缓慢的问题,通过将领域知识引入到SEM算法中,提出了KB-SEM算法,该算法首先用D-S证据理论综合领域知识,然后将采集的知识以禁忌表的方式嵌入SEM中来限制和引导算法的搜索路径,缩小算法的搜索空间。实验表明,KB-SEM算法能有效地提高算法的学习精度和时间性能,且能在一定程度上避免主观偏见和数据噪音的干扰。

关键词: 贝叶斯网络, 领域知识, 缺省数据, KB-SEM

Abstract: To overcome shortcomings present in the SEM in the presence of missing values data sets such as lower precision and slow convergent speed,KB-SEM is proposed by introducing domain knowledge.A tabu list derived from collecting domain expert knowledge synthesized by using D-S evidence theory is embedded in the SEM to constrain and guide the searching path,and to decrease the searching space.Experimental result shows that KB-SEM can improve the precision and the speed effectively,and to some extent avoid subjective bias and disturbance of noise in the data sets.

Key words: Bayesian Networks(BNs), domain knowledge, missing data, KB-SEM