Fuzzy Cognitive Maps: A Business Intelligence Discussion

Abstract

Modeling complex systems by means of computational models has enabled experts to understand the problem domain without the need of waiting for the real events to happen. In that regard, fuzzy cognitive maps (FCMs) have become an important tool in the neural computing field because of their flexibility and transparency. However, obtaining a model able to align its dynamical behavior with the problem domain is not always trivial. In this paper, we discuss some aspects to be considered when designing FCM-based simulation models by relying on a business intelligence case study. In a nutshell, when the fixed point is unique, we recommend to focus on the number of iterations to converge instead of focusing on the reached attractor and stress the importance of the transfer function chosen in the model.