Evolutionary Multilayered Fuzzy Cognitive Maps: A Hybrid System Design to Handle Large-Scale, Complex, Real-World Problems

This paper proposes an extension to multilayered fuzzy cognitive maps (ML-FCMs) and introduces a new methodology based on ML-FCMs aiming at enhancing their capabilities for scenario analysis and forecasting. The main issue here is the decomposition of the parameters into smaller, more manageable quantities, organised in a hierarchical structure forming a model, which consists of subsystems working together and supporting a central objective. The modelling of a particular large scale system is primarily represented by a main, central FCM, with distinct sub-models (layers) implemented also as FCMs and linked together in a hierarchical tree structure. The sub-models represent and implement (in computational terms) the decomposed parameters and variables of the system, thus offering the ability of isolating and studying critical parts of the system. The objective of the evolutionary multilayered FCM approach, as it is proposed in this work, is to improve the decision-making process of basic ML-FCMs by integrating a genetic algorithm (GA) for the production of a set of solutions in the form of new weight matrices for any targeted activation level throughout the multilayered structure