Learning fuzzy cognitive maps using Structure Optimization Genetic Algorithm

Fuzzy cognitive map (FCM) is a soft computing methodology that allows to describe the analyzed problem as a set of nodes (concepts) and connections (links) between them. In this paper a new Structure Optimization Genetic Algorithm (SOGA) for FCMs learning is presented for modeling complex decision support systems. The proposed approach allows to automatic construct and optimize the FCM model on the basis of historical multivariate time series. The SOGA defines a new learning error function with an additional penalty for highly complexity of FCM understood as a large number of concepts and a large number of connections between them. The aim of this study is the analysis of usefulness of the Structure Optimization Genetic Algorithm for fuzzy cognitive maps learning. Comparative analysis of the SOGA with other well-known FCM learning algorithms (Real-Coded Genetic Algorithm and Multi-Step Gradient Method) was performed on the example of prediction of rented bikes count. Simulations were done with the ISEMK (Intelligent Expert System based on Cognitive Maps) software tool. The obtained results show that the use of SOGA allows to significantly reduce the structure of the FCM model by selecting the most important concepts and connections between them.