Hybrid Module of the IACVIRTUAL Project

Abstract

This work represents the hybrid module of the IACVIRTUAL meta-environment. In this context we will basically approach the Hybrid Expert System (HES), which is composed by the Neural Networks Based Expert System (NNES) and by the Rule-Based Expert System (RBES). The HES is destined to support the decision of a clinical-surgical team, in the area of cardiology, in the definition of a therapeutic conduct in patients with coronary heart disease. The implementation process starts with the Knowledge Acquisition (KA), which comes from the analysis of a series of clinical parameters, which are used as input data for the NNES. This way, knowledge acquired during elicitation is converted in fuzzy rules. Through these rules, elicitated knowledge is mapped in AND/OR graphs, which then represent the starting structure of the NNES. Learning and optimization of the RBES are made through the Genetic-Backpropagation Based Learning Algorithm (GENBACK). This algorithm can, during the learning process, modify the weight of the connections, as well as the network structure. Knowledge abstracted from the RBES, being already refined, as well as trained and tested, is used to form the Knowledge Base of the RBES.