An Approach for Representing and Inferring Uncertainties among Attributes and Classes in Multidimensional Data
Mingliang Yue, Kun Yue, Weiyi Liu · 2010
It is desirable to represent and infer uncertain knowledge under multidimensional situations due to the wide applications of multidimensional data. Bayesian network (BN) is a generally accepted framework for representing and inferring probabilistic causalities among random variables. In this paper, by adopting the notation of classification, we use an extended augmented naive Bayesian network, called EAN, to represent and infer uncertainties among attributes and classes in multidimensional data. We present the methods for constructing and inferring of uncertainties with an EAN, by extending those on general BNs. Based on EAN, the inference of uncertainties can be done among classes and attributes under arbitrary evidences. Experiments results verify the feasibility and effectiveness of our methods.