Optimising retrieval phase in CBR through Pearl and JLO algorithms for medical diagnosis

Abstract

Case-based reasoning CBR is an approach of solving problem, which is based on the reuse, by analogy, of past experiences called 'case'. Retrieval of cases is a primary step in CBR, and the similarity measure plays a very important role in case retrieval. We present a probabilistic retrieval phase applied to the diagnosis of hepatic pathologies. The main idea consists in modelling the case base by a Bayesian network BN. BN is an excellent tool for modelling the uncertainty in terms of their clear graphic representation as well as the conditional probabilities laws defined on. In this paper, a retrieval phase is attended, which consists of selecting the most similar case of log-linear model by considering Bayesian network as a log-linear model on the simplification of the probability. Also, we used two exact algorithms of inference: Pearl and JLO-Jensen, Lauritzen and Olesen - to calculate the conditional probabilities and to optimise the retrieval phase. The objective of this work is to improve the performance of CBR. Experimental results show that the proposed methods improve the efficiency of case retrieval.