A possibilistic framework for single-fault causal diagnosis under uncertainty
Résumé
This paper presents a general approach to diagnosis in a relational setting where uncertainty is expressed by means of possibility theory. Causal knowledge is supposed to be described by directed links between causes and their possible symptoms. More precisely, symptoms are described on binary or non-binary attribute domains by means of fuzzy sets and may be pervaded with uncertainty. Moreover, observations may be also fuzzy and uncertain. The proposed model generalizes both a previously developed approach to binary symptoms on the one hand, and a treatment of graded symptoms in the case of precise observations also recently introduced on the other hand.