Similarity of fuzzy data in a case-based fuzzy system in anaesthesia

Abstract

Real world expert systems generally require a large amount of explicit deep knowledge about the problem domain. Case-based expert systems can provide a way of generating solutions even without a complete model of the problem's domain. Based on stored data representing implicit knowledge, the search for similar problem situations and the adaptation of their solutions to the given problem is the central idea of case-based reasoning. A formal definition of case-based reasoning is proposed integrating fuzzy similarity analysis, indexing, combination and modification of data. Taking postoperative pain management as a real world example from a medical problem domain a method for generating solutions from fuzzy data is shown. This operational process of calculating similarities transforms data or better objects to similar cases. Well known AI-techniques like classification or rule-based methods can be integrated for refinement and learning to achieve better solutions in the course of time. In the sense of soft computing computer-assisted solutions are generated not necessarily optimal or best ones but often acceptable ones. Finally this method gives the opportunity to develop an expert system shell usable for many different problem domains.