ON THE IMPROVEMENT OF THE METHOD OF FUZZY INFERENCE IN KNOWLEDGE-BASED SYSTEMS

Abstract

The article discusses two methods of developing and researching fuzzy inference systems for diagnosing a patient's
condition based on a set of fuzzy symptoms. The problem of determining the patient's disease according to the data of the
diagnosis of the state of health and current symptoms is described. To solve this problem in the context of fuzzy logic, a system
of fuzzy logical inference is proposed, which uses the information of dependencies between diseases, symptoms, and levels of
symptoms. A table of correspondence between the levels of specific symptoms and the disease is presented. Fuzzy modified
triangular membership functions were constructed for each linguistic variable of current symptom and disease level. Several
ways of setting rules are considered, which are outlined within the framework of two methods. The first method uses a standard
approach to building a fuzzy inference system. It is based on the concept of a linguistic variable and a fuzzy rule, which is
created on the basis of expert knowledge. Intuitive direct rules are formed, which are presented in mathematical form. Using
this rule system, various derivation mechanisms can be applied, such as Mamdani derivation. The second method, in turn,
presents an improved fuzzy model that includes additional linguistic variables and rules to quantify the correctly specified
symptoms in the set of all symptoms of the patient's disease. This system of rules operates on linguistic variables that use the
number of correctly given symptoms relative to the total number of symptoms of the disease. A comparison of the described
logical inference systems was made on the test data set. The proposed methods are analyzed for their effectiveness and accuracy
in establishing a diagnosis. The results of the study may be important for the development of new diagnostic methods based on
fuzzy logic, which may prove to be more effective compared to traditional methods.