Decision support systems are effectively used in various areas of practical activity. However, there is no noticeable progress in the application of such systems in crime investigation, although solutions for the formation of knowledge bases as the basis for decision support systems are diverse. As a rule, the problem of understanding the logic of the decision support system by the decision maker is ignored. The transition from databases (datasets) to knowledge bases can be assessed as the most important tool to achieve the criterion of understandability of the operation of decision support systems. This paper proposes a method for constructing a forensic knowledge base based on a fuzzy neural network. An inference system has been developed that allows for assessing a suspect’s predisposition to commit a crime. An approach to structuring forensic crime characteristics is developed. The structure of a forensic crime characteristic contains three components: element, feature, and gradations. The number of gradations for each input and output feature varies and depends on the forensic meaning of the feature.