Shell para desenvolvimento de sistemas especialistas na área de saúde

Abstract

The expert systems are computational tools built from argument of strong methods in artificial intelligence. This argument states that the resolution of complex problems is only possible when specific knowledge of the nature of the problem to be solved is known. The construction of expert systems in addition to be a very complex task involves a precious expenditure of time. Due to this, there are tolls called Shell that allow rapid prototyping of an expert system, demonstrating or not the viability of a Integral System construction. In this work, the main concern was to develop a Shell that was used with priority, but not exclusively, to diagnosis applications in medicine. In this sense, the system has been provided with the ability to handle linguistic data, data like image and numeric interval data (expressed by ranges of values: a x b). The possibility of working with numerical interval data and with multivalued variables is the main difference between Shell proposed in this work compared to other publications found in literature. In order to be possible to deal with interval data, a specific inference algorithm that incorporates an unification algorithm was developed. It has been given to the system the treatment of uncertainty through certainty factors. There is the possibility of working with multiples targets simultaneously. In the results, an original system was developed for atherosclerosis diagnostics that uses rules with interval variables and a system for AIDS treatment that uses multivalued variables. Moreover, the uncertainty treatment in the Shell proposed is compared with the uncertainty treatment in two other systems nationally developed. As a conclusion, the great application of the proposed system is highlighted, which allows the diagnosis in areas so varied that ranges from radiology to laboratory diagnosis.