Bayesian Metanetworks for Mobile Web Content Personalization

The problem of profiling and filtering the Web content is important particularly for mobile applications where wireless network traffic and mobile terminal size are limited comparing to the Internet access from the PC. Bayesian networks are known to be good tool for learning user preferences in electronic commerce. However more sophisticated cases, when user preferences are changing according to a context or location changes, require also more sophisticated Bayesian architectures. In this paper Multilevel Bayesian belief networks were proposed for modeling of user preferences in the mobile information environment. Bayesian Metanetwork is a multilevel probabilistic model which can be used for prediction of mobile user’s preferences taking into account special requirements of mobile environment. In this paper we consider several reasonable modifications of Bayesian Metanetwork, appropriate reasoning rules with some discussion about possible implementation. Further research is expected to specify algorithms for learning Bayesian Metanetworks and to define application area more concrete.