Analysis of an evolutionary algorithm for complex fuzzy cognitive map learning based on graph theory metrics and output concepts

The fuzzy cognitive map (FCM) is an effective tool for modeling dynamic decision support systems. It describes the analyzed phenomenon in the form of key concepts and the causal connections between them. The main aspects of the building of the FCM model are: concepts selection, determining the output concepts, criterion selection, and determining the relationships between concepts. It is usually based on expert knowledge. The main goal of the paper is to define the optimal in some sense FCM structure through the introduction of the notion of output concepts and minimizing the number of concepts and connections between them. The proposed approach allows for: (1) the selection of key concepts based on graph theory metrics and determining the connections between them; (2) the determination of the criterion of learning based on output concepts and fitting the learning process to the analyzed problem. A simulation analysis was done with the use of synthetic and real-life data. Experiments confirm that the proposed approach improves the learning process compared to the standard approaches.