Research on the Product Configuration Method Based on Constraint Satisfaction Problem and Bayesian Network

Zhiqiang Liu, Yila Su, Huimin Li, Fei Wang
2015 International Journal of u- and e- Service, Science and Technology  
In the context of mass customization, to address the issues about method based on Constraint Satisfaction Problem (CSP) in the field of product configuration, a combined product configuration method based on the CSP and Bayesian Network (BN) is proposed in this paper. On the basis of CSP and BN theory, a product configuration model established based on CSP and BN, and the specific method for solving was given, Including the reasoning on posterior probability by establishing the Bayesian Network
more » ... model and the reasoning on Constraint Satisfaction Problem. Finally, to assemble the computer as an example, a configuration system of assembly computer product is developed so as to verify the feasibility and effectiveness. The user's preference is reasoned by Bayesian network model, the modeling process, including the Bayesian network structure learning and parameter learning, which is the conditional probability of network nodes learning. Based on the characteristics of the field of product configuration, the paper uses the logical structure of the product as the Bayesian network topology directly, for example, to assemble the computer shown in Figure 1 . The structure is a tree, a special Bayesian network structure, the logical structure nodes of the product is used as Bayesian network nodes, which are also set as the user's preference, directed edges show the relationships between the nodes intuitively. 292 Copyright ⓒ 2015 SERSC supports exact and approximate reasoning, but also supports topology learning and parameter learning as well as static and dynamic models. After learning through structural BNT shown in Figure 2 , contains 37 nodes, each variable value is {1, 2}, reflects the directed edges between nodes. Figure 2. The Bayesian Network Structure of Assembled Computer in BNT In BNT, this article will traindata.txt data samples into a text file as the training sample, each act of a user configuration examples, with 1 and 2 that includes all the nodes in the recording format, comma between nodes separated. And the learning method is Bayesian estimation method, in BNT learning, the statement is bayes_update_params (). Determine the structure and parameters of Bayesian network that determines Bayesian network model, then you can use this model to do probabilistic reasoning.
doi:10.14257/ijunesst.2015.8.4.26 fatcat:l2tvy6jecvfzxa5uhcfydsg3ru