[1]王双成,李小琳,侯彩虹.用于因果分析的混合贝叶斯网络结构学习[J].智能系统学报,2007,2(6):82-90.
WANG Shuang-cheng,LI Xiao-lin,HOU Cai-hong.Learning in a hybrid Bayesian network structure for causal analysis[J].CAAI Transactions on Intelligent Systems,2007,2(6):82-90.
点击复制
WANG Shuang-cheng,LI Xiao-lin,HOU Cai-hong.Learning in a hybrid Bayesian network structure for causal analysis[J].CAAI Transactions on Intelligent Systems,2007,2(6):82-90.
用于因果分析的混合贝叶斯网络结构学习
《智能系统学报》[ISSN 1673-4785/CN 23-1538/TP] 卷:
2
期数:
2007年第6期
页码:
82-90
栏目:
学术论文—机器学习
出版日期:
2007-12-25
- Title:
- Learning in a hybrid Bayesian network structure for causal analysis
- 文章编号:
- 1673-4785(2007)06-0082-08
- 分类号:
- TP18
- 文献标志码:
- A
- 摘要:
- 目前主要结合扩展的熵离散化方法和打分—搜索方法进行混合贝叶斯网络结构学习，算法效率和可靠性低，而且易于陷入局部最优结构.针对问题建立了一种新的混合贝叶斯网络结构迭代学习方法.在迭代中，基于父结点结构和Gibbs sampling进行混合数据聚类，实现对连续变量的离散化，再结合贝叶斯网络结构优化调整，使贝叶斯网络结构序列逐渐趋于稳定，可避免使用扩展的熵离散化和打分—搜索所带来的主要问题.
- Abstract:
- At present, learning in a hybrid Bayesian network structure mainly depends on a combination of the searching & scoring method and the expanded entropy discretiz ation algorithm. However, the algorithm is prone to fall into local optimal trap s and its efficiency and reliability are not good. In this paper, a new iterativ e method for learning with hybrid Bayesian network structures is presented. In e ach iteration, mixed data clustering is carried out based on father mode structu res and Gibbs sampling, so that continuous variables are discretized. Then, thro ugh optimization of the Bayesian network structure, the sequence of Bayesian net work structures gradually tends to converge, avoiding the main problems encounte red with the expanded entropy discretization algorithm.
- 参考文献/References:
-
［1］ PEARL J. Probabilistic reasoning in intelligent systems: networks of plaus ible inference［M］. San Mateo, Morgan Kaufmann, 1988.
［2］HECKERMAN D, GEIGER D, CHICKERING D M. Learning Bayesian networks: the com bi nation of knowledge and statistical data［J］. Machine Learning, 1995, 20(3): 19 7-243.
［3］SPIRTES P, MEEK C, RICHARDSON T. Causal inference in the presence of laten t variables and selection bias［A］. Proceedings of the 11th Annual Conference on Uncertainty in Artificial Intelligence［C］. Pittsburgh, USA, 1995.
［4］CHICKERING D M. Learning equivalence classes of Bayesian network structure s［J］. Machine Learning, 2002, 2(3): 445-498.
［5］HENSON J. Comparing causality principles［J］. Studies in History and Phil osophy of Modern Physics, 2005, 36(3): 519-543.
［6］THIESSON B, MEEK C, CHICKERING D, HECKERMAN D. Learning mixtures of Bayesi an networks［R］. MSRTR9730, 1997.
［7］MURPHY K P. Inference and learning in hybrid Bayesian networks［R］. CSD 98990,1998.
［8］MONTI S, COOPER G F. learning hybrid Bayesian networks from data［R］. ISSP 9701, 1997.
［9］FAYYAD U, IRANI K. Multinterval discretization of continuousvalued att ribu tes for calssification learning［A］. Proceedings International Joint Confer ence on Artificial Intelligence［C］. Chambery, France, 1993.
［10］LAM W, BACCHUS F. Learning Bayesian belief networks: an approach base d on the MDL principle［J］. Computational Intelligence， 1994, 10(4): 269-293.
［11］CHICKERING D M. Learning Bayesian networks is NPHard［R］. MSRTR94 17, 1994.
［12］茆诗松,王静龙,濮晓龙.高等数理统计［M］. 北京： 高等教育出版社, 1998 .
?［13］GEMAN S, GEMAN D. Stochastic relaxation, Gibbs distributions and t he Baye si an restoration of images［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1984, 6(6): 721-742.
［14］CHOW C K, LIU C N. Approximating discrete probability distributions with de pendence trees［J］. IEEE Transactions on Information Theory, 1968, 14(3): 462-4 67.
［15］BUNTINE W L. Chain graphs for learning［A］. Proceedings of the 17th Conference Artificial Intelligence［C］. San Francisco, USA, 1995.
［16］DOMINGOS P, PAZZANI M. On the optimality of the simple Bayesian classifie r under zeroone loss［J］. Machine Learning, 1997, 29(2-3): 103-130.
［17］王双成,苑森淼.具有丢失数据的贝叶斯网络结构学习研究［J］.软件学报,20 04, 15(7): 1030-1041.
?WANG Shuangcheng, YUAN Senmiao. Research on learning Bayesian networks structu re with missi ng data［J］. Journal of Software, 2004, 15(7): 1030-1041.
［18］王双成,苑森淼.具有丢失数据的可分解马尔科夫网络结构学习［J］. 计算机学报, 2004, 27(9): 1221-1228.
?WANG Shuangcheng, YUAN Senmiao. Learning decomposable Markov network structure with missing data［J］. Chinese Journal of Computers, 2004, 27(9): 1221-1228.
［19］王 飞,刘大有,薛万欣.基于遗传算法的Bayesian网中连续变量离散化的研究［J］. 计算机学报, 2002, 25(8): 794-800.
WANG Fei, LIU Dayou, XUE Wanxin. Discretizing continuous variables of Bayesia n network s based on genetic algorithms［J］. Chinese Journal of Computers, 2002, 25(8): 7 94-800.
［20］MURPHY S L, AHA D W. UCI repository of machine learning databases［ EB/OL］. http: //www. ics. uci. edu/~mlearn/MLRepository, 2005-09-10
- 相似文献/References:
-
[1]花黄伟,杨春燕.基于因果分析的事元蕴含系及其应用[J].智能系统学报,2017,12(1):60.[doi:10.11992/tis.201610012]
HUA Huangwei,YANG Chunyan.Affair-elements implication system andits application based on causal analysis[J].CAAI Transactions on Intelligent Systems,2017,12():60.[doi:10.11992/tis.201610012]
备注/Memo
收稿日期：2007-01-04.
基金项目：
国家自然科学基金资助项目(60675036)；
上海市重点学科基金资助项目(P1601)；
上海市教委重点基金资助项目(05zz66).
作者简介：
?王双成，男，1958年生，教授，博士，主要研究方向为人工智能、机器学习和数据挖掘及其在风险管理中的应用，先后承担国家自然科学基金“面向智能信息处理的贝叶斯网络关键理论与方法研究”和“面向风险管理的贝叶斯网络与集成研究”等课题，发表学术论文 40 余篇. E-mail: wangsc@lixin.edu.cn.
李小琳，女，1978年生，讲师，博士，主要研究方向为机器学习和数据挖掘，先后参与国家自然科学基金2项，发表学术论文16篇. E-mail: lixl_126@126.com.
侯彩虹，女，1978年生，讲师，博士，主要研究方向是智能控制和数据挖掘，先后参与国家自然科学基金2项，发表学术论文12篇.E-mail: dhuhch@mail.dhu.edu.cn.
更新日期/Last Update:
2009-05-09