Bayesian Network-Based Service Context Recognition Model

Service context refers to the environmental information which influences the service execution. The processing of service context information is the foundation of the future intelligent telecommunication services in the ubiquitous convergent network. As the complexity and variability of the real world, there exists a large number of uncertain service context information, such as the imprecise information collected by sensors, data noise, and inaccurate location by different location technologies. Hence, it is crucial to recognize the correct service context environment based on unreliable context information. In recent years, context computing and context-awareness have become a major topic of research in an ubiquitous computing field. Ontology technology is often used to support context modeling and reasoning. And the probability theory, especially the Bayesian network, is adopted to deal with uncertainty. However, much of the existing relevant research work mainly concentrated on the modeling and reasoning of context information. But the way to construct and evolve the service context recognition model supporting uncertain reasoning effectively and systematically has not involved. In addition, most of the relative work is only the qualitative analysis, and lacks the quantitative performance analysis and experimental verification. What is more, the research on how to apply context-awareness technology to the telecommunication field and support the intelligence and individualization of the service is scarce. In this article, we combine context-awareness technology with telecommunication service network, and discuss the mechanism of service intelligence in ubiquitous convergent network environment. Then a construction method of the service context recognition model based on the Bayesian Network is put forward. The process of constructing the Bayesian network-based service context recognition model is mainly divided into the following steps: First, according to different service requirements and scenarios, the specific context problem domain should be specified; second, constructing the service context recognition model, which includes the determination of the topology structure and the node probability distribution of Bayesian network; third, when the preliminary Bayesian network-based service context recognition model is constructed, it could be initially applied to the practical system to do the clustering or causality analysis; and last, when the model cannot satisfy the need, the model update should be performed based on the feedback information of the evaluation step and the newly-generated plentiful sample data. This approach has been applied to intelligent access adaptation processing of call service in the office environment. The experimental results show that this approach can strengthen the intelligence of service and compensate for restrictions of certain reasoning.