Advanced Bayesian Network Models with Fuzzy Extension

Abstract

Fuzzy Bayesian networks (FBNs) are the variant of standard/classical Bayesian networks (BNs), which have intrinsic capability of handling ambiguity due to lack of expert knowledge and eventually reduces the epistemic uncertainty when used as computational models. Of late, FBNs have gained substantial research interest to be applied for time series prediction in both non-spatial and spatial domains . This chapter discusses a number of fuzzy BN models that have recently been proposed in literature. The central attention is paid on how the discrete Bayesian analysis in the previously discussed enhanced BN models can be further improved through incorporated fuzziness so as to make them more realistic for dealing with various contexts of spatial time series prediction. A comparative study, at the end of the chapter, demonstrates superiority of the fuzzified enhanced BN models, compared to those having no incorporated fuzziness.