An Efficient Algorithm for Naive Possibilistic Classifiers with Uncertain Inputs

Abstract

Possibilistic networks are graphical models particularly suitable for representing and reasoning with uncertain and incomplete information. According to the underlying interpretation of possibilistic scales, possibilistic networks are either quantitative or qualitative. In this paper, we address possibilistic-based classification with uncertain inputs. More precisely, we first analyze Jeffrey's rule for revising possibility distributions by uncertain observations. Then, we propose an efficient algorithm for revising possibility distributions encoded by a naive possibilistic network. This algorithm is particularly suitable for classification with uncertain inputs since it allows classification in polynomial time using different efficient transformations of initial naive possibilistic networks.