Application of emotional learning fuzzy inference systems and locally linear neuro-fuzzy models for prediction and simulation in dynamic systems

Abstract

Mathematical description and modeling of dynamic systems is challenging due to their high level of complexity, their nonlinear and chaotic behaviors, the presence of uncertainties and interference of human behavior in their outputs, and their time-variant nature. Because of such characteristics and the importance of dynamic systems modeling, high-performance modeling tools are required to analyze, identify, model and finally control such systems. Emotional learning fuzzy inference system (ELFIS) and locally linear neuro-fuzzy (LLNF) model can be considered as two potential tools for modeling and prediction of dynamic systems. In this paper ELFIS and LLNF are applied to three various dynamic systems, namely electricity price forecasting in competitive power markets, stock market prediction and prediction of surface ozone concentration. the comparisons between the applied methods (LLNF and ELFIS) and some other methods such as multi-layer perceptron (MLP) neural networks, demonstrated the superiority and computational efficiency of the proposed approaches over the other methods, besides their greater comprehensibility and transparency for dynamic systems modeling and prediction.