| 小数据集的贝叶斯网络结构学习 |
| |
| 引用本文： | 王双成,冷翠平,李小琳.小数据集的贝叶斯网络结构学习[J].自动化学报,2009,35(8):1063-1070. |
| |
| 作者姓名： | 王双成 冷翠平 李小琳 |
| |
| 作者单位： | 1.上海立信会计学院数学与信息学院 上海 201620 |
| |
| 基金项目： | 国家自然科学基金(60675036,60803055);;上海市教委重点学科基金和上海市教委科研创新重点项目(09zz202)资助~~ |
| |
| 摘 要： | 针对直接基于小数据集贝叶斯网络结构学习不可靠, 以及目前对小数据集的处理只强调扩展而忽略对扩展数据的修正等, 提出了将扩展与修正相结合的小数据集处理机制, 以及在此基础上的基于结点排序和局部打分--搜索的贝叶斯网络结构学习方法. 可不需要完全结点顺序的先验知识, 但能够结合专家的部分结点顺序信息. 实验结果显示了这种方法的有效性和可靠性.
|
| 关 键 词： | 贝叶斯网络 小数据集 结构学习 最大似然树 吉布斯抽样 |
| 收稿时间： | 2008-6-23 |
| 修稿时间： | 2008-12-31 |
Learning Bayesian Network Structure from Small Data Set |
| |
| WANG Shuang-Cheng LENG Cui-Ping LI Xiao-Lin .School of Mathematics , Information,Shanghai Lixin University of Commerce,Shanghai .Opening Economy , Trade Research Center,Shanghai .School of Business,Nanjing University,Nanjing.Learning Bayesian Network Structure from Small Data Set[J].Acta Automatica Sinica,2009,35(8):1063-1070. |
| |
| Authors: | WANG Shuang-Cheng LENG Cui-Ping LI Xiao-Lin School of Mathematics Information Shanghai Lixin University of Commerce Shanghai Opening Economy Trade Research Center Shanghai School of Business Nanjing University Nanjing |
| |
| Affiliation: | 1.School of Mathematics and Information, Shanghai Lixin University of Commerce, Shanghai 201620;2.Opening Economy and Trade Research Center, Shanghai Lixin University of Commerce, Shanghai 201620;3.School of Business, Nanjing University, Nanjing 210093 |
| |
| Abstract: | It is incredible to learn Bayesian network structure directly from small data set. For improving the reliability, many methods of extending small data set have been developed, but the revision of extended data is neglected. In this paper, extending small data set is combined with revising extended data to upswing the data reliability. A directed tree is built from the small data set and variables are sorted according to it. On the basis of the variable order, a Bayesian network structure can be established based on the local search and scoring method. This method dose not need the prior knowledge of the variable order, but the partial order information of expert can be used properly. Experimental results show that this method can effectively learn Bayesian network structure from a small data set. |
| |
| Keywords: | Bayesian network small data set structure learning maximal likelihood tree Gibbs sampling |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《自动化学报》浏览原始摘要信息 |
|
点击此处可从《自动化学报》下载全文 |
|
