Volume 15, Number 2 (2016) | JIRSS 2016, 15(2): 45-61 | Back to browse issues page



DOI: 10.18869/acadpub.jirss.15.2.45

XML Persian Abstract Print


Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:

Rezaei Tabar V, Mahdavi M, Heidari ‎, Naghizadeh S. Learning Bayesian Network Structure Using Genetic Algorithm with Consideration of the Node Ordering via Principal Component Analysis. JIRSS. 2016; 15 (2) :45-61
URL: http://jirss.irstat.ir/article-1-304-en.html

Department of Statistics‎, ‎Faculty of Mathematics and Computer Sciences‎, ‎Allameh Tabataba'i University‎, ‎Tehran‎, ‎Iran and School of Biological sciences‎, ‎Institute for Research in Fundamental Science (IPM)‎, ‎Tehran‎, ‎Iran
Abstract:   (1190 Views)

The most challenging task in dealing with Bayesian networks is learning their structureTwo classical approaches are often used for learning Bayesian network structure; Constraint-Based method and Score-and-Search-Based one. But neither the first nor the second one are completely satisfactory. Therefore the heuristic search such as Genetic Algorithms with a fitness score function is considered for learning Bayesian network structure. To assure the closeness of the genetic operators, the ordering among variables (nodes) must be determined.

In this paper, we determine the  node ordering by considering the Principal Component Analysis (PCA). For this purpose we first determine the appropriate correlation between variables and then use the absolute value of variable's coefficients in the first component. It means that a node X_i can only have the node X_j as a parent, if the absolute value of   coefficient X_j in first component will be higher than X_iWe then use the Genetic Algorithm with fitness score BIC regarding the  node ordering  to construct the Bayesian Network. Experimental results over well-known networks  Asia, Alarm and Hailfinder  show that our new technique  has higher accuracy and better degree of data matchingIn additionwe apply our technique to the  real data set which is related to Bank's debtor that owe over 500 million Rials to Bank Maskan (Housing Bank) in Iran. Results also show that the proposed technique has greater modeling power than other node ordering techniques such as   Hruschka et al. (2007), Chen et al. (2008) and K2 algorithm.

Full-Text [PDF 441 kb]   (252 Downloads)    
Type of Study: Original Paper |
Received: 2015/04/7 | Accepted: 2016/08/25 | Published: 2016/08/25

Add your comments about this article : Your username or email:
Write the security code in the box

Send email to the article author


© 2015 All Rights Reserved | Journal of The Iranian Statistical Society

Designed & Developed by : Yektaweb