Flow Graph Approach for Studying Fuzzy Inference Systems

This paper proposes an approach to analysis of fuzzy inference systems. To this end, a flow graph representation of a decision table with linguistic values is used. Every layer in a fuzzy flow graph corresponds to a fuzzy attribute (input or output). Each node in a layer represents a particular linguistic value of an attribute. Certainty factor, coverage factor and strength of rules can be computed for a given set of examples (universe) and serve as helpful measures for gaining more insight into the operation of a fuzzy inference system. Moreover, application of flow graphs can be useful for selecting membership functions and determining other parameters in the design of fuzzy inference systems.