On knowledge representation in fish disease diagnosis expert system

Abstract

Knowledge representation is the generator in Fish Disease Diagnosis Expert System (Fish-expert), so it is necessary that develop an explicit formalization capable of compiling the knowledge of experts to uniformity of expression and describing all kinds of fish diseases. In this paper we focus on the knowledge representation formalism in Fish-expert. With integrated fish disease diagnostics with experiences from many experts had accumulated through many years, we brought forward the conceptualization architecture for fish disease diagnosis, which describes the causality relations inherent in fish disease diagnosis knowledge. Then we represented them via a hybrid formalism that combines conventional and logic production rules and Neural Network. The conventional rule is employed to represent the knowledge from experience of experts cutting down the problem space. The logic rule represents the causality relations among symptoms, diseases and pathogenies and The Neural Network is used for training and forming the relations among them. Compared with the knowledge representation formalism only employed the conventional rule, the fundamental characteristic of the formalism is that it can represent both the frequency of occurrence of symptoms with diseases and the strength of confirmation of symptoms for diseases. With application of the hybrid formalism the amount of rules of Fish-expert had been decreased from 1890 to 312 and the reasoning speed and veracity had been promoted from 1 second to 0.5 second and 85% to 95% or more separately.