The Model Of Distributed Heterogeneous Bayesian Network And Its Application Of Mobile Commerce

As wireless electronic and information technology-based mobile telecommunications industry developping rapidly, E-Commerce has not faded out, Mobile commerce has emerged into our vision. It has some advantages such as highly-efficient, time-saving, convenient and personalized, which traditional E-commerce can not match with. However, "China's mobile value-added services market analysis report (2005)" pointed out that 32 percents of the mobile users are unsatisfied with mobile message service, of which 54 percents are unsatisfied for that the service is not personal. Therefore, how practical realization of the mobile users personalized recommendation becomes one of the urgent problems of mobile business.However, the mobile commerce along the various restrictions on their own personalized recommendation has brought many problems, including distributed heterogeneous datas, the high cost of communications, limited bandwidth, protection of privacy, as well as the low level of flexibility poor site structures. Therefore, how to use the appropriate personalized recommendation technology address these problems so as to establish a personalized recommendation system becomes the focus of this paper.In recent years, the bayesian network model as one of the personalized recommendation technologies, with its theoretical strict and consistency, and effective local computer system and intuitive graphical knowledge representation, becomes the hotspot of personalized recommendation technology fields. However, the most existing researches on bayesian network only confined to the centralized method, therefore, they can not be the solution to the above-mentioned problems.This paper fully absorbed previous research results, proposed a perspective that using distributed heterogeneous bayesian network model to construct personalized recommendation model of mobile commerce, and given the relevant learning algorithm, thereby solving the above-mentioned problems, built the model accuratually and rapidly.The content and results this paper including:(1) Bayesian network structure learning algorithm based on CI test. Make the compare between the mutual information between the nodes and the threshold as CI test standards, propose the LBNSL algorithm and analyses its complexity. Analysis of results showed the bayesian network structure was accurate and the algorithm complexity is not high that used the LBNSL algorithm.(2) Distributed learning of bayesian network. Make the compare between probability of sample and the threshold as the standards of how to transfer samples, propose the DHBN algorithm and the comparative analysis method between two models using the theory of relative entropy.(3) Distributed heterogeneous bayesian network modeling study and the analysis of model. Based on the learning algorithm and comparative analysis method proposed in this paper, the paper have studied the whole process of modeling the distributed heterogeneous bayesian network and analyzed the modeling results between using proposed method and traditional method. The results show that the structures are accordable with a little difference on parameters.(4) The application of distributed heterogeneous bayesian network model in mobile commerce. Propose the structure and function of mobile commerce personalized recommendation. Then in the realization of system, make Hugin as thesoftware platform, make the reasoning of built distributed heterogeneous bayesian network model as a result of personalized recommendation to mobile users, thereby system realizes the application of function.