Use of Fuzzy Interval Risk Assessments in Decision Making in Predictive Diagnostics

Expert systems accumulate the experience of professionals when solving practical problems in loosely structured areas, for example, in predictive diagnostics. The decisions formed by the expert system model the thinking process of a professional. When describing subject areas that are characterized by the presence of uncertainty factors and a large number of qualitative parameters, it is advisable to use fuzzy models. At the same time, when solving forecasting problems, it is appropriate to use the principles of situational control, when a lot of typical fuzzy situations are used, the analysis of proximity to which makes it possible to predict the occurrence of emergency situations. The process of knowledge extraction, as a necessary stage in the formation of the knowledge base of an expert system, affects the efficiency of the development of an expert system and the quality of its functioning. A significant issue is the problem of reducing the process of extracting knowledge when communicative methods are used. Communicative methods of acquiring knowledge involve the formation of a knowledge representation model based on information obtained as a result of interviewing experts. For example, in medicine, the decisions made depend on many features, each of which can take on a large number of values. The combination of a large number of values leads to the need to generate risk assessments for a significant number of situations, which is the reason for the high complexity. The article substantiates a method that makes it possible to reduce the complexity of the process of interaction with an expert by changing the type of questions presented to experts, which makes it possible to establish in the course of the survey not absolute, but interval assessments of the risk of emergency situations for classes of typical situations. This makes it possible to obtain an acceptable for practice labor input for preparing the knowledge base of an expert system, which linearly depends on the number of features and their values.