Facing multicollinearity in a Fuzzy Cognitive Maps environment using bias nodes

Abstract

This paper presents an improvement in the Fuzzy Cognitive Maps (FCMs) theory used in modelling dynamical systems. The dynamic behaviour of a system focuses on a stable behaviour reaching equilibrium state at fixed points. Multicollinearity in FCMs, is examined as a phenomenon possibly present in any dynamic modelling. Cases in which the activation level of a concept is constantly driven towards +1 or -1 are investigated. In addition, concepts stabilizing at levels other than those reflecting the actual environment are also considered. These experiments are performed by selectively inserting bias nodes in the map so as to increase the reliability of the dynamic model. Following the theoretical background of FCMs and the mathematical formulation of bias nodes we demonstrate their use and contribution in terms of a case study that deals with the Cyprus Issue in view of the Turkish full EU membership.