An Interval Type 2 Fuzzy Extreme Learning Machine-Based Brain Emotional Learning

Over the past three decades, Brain Emotional Learning (BEL) models, inspired by the neural architecture of the mammalian limbic system, have been established as effective machine intelligence frameworks capable of integrating artificial neural networks and neuro-fuzzy systems. This study introduces a novel neuro-fuzzy prediction model, BELIT2FELM, which combines BEL with an Interval Type 2 Fuzzy Extreme Learning Machine (IT2FELM) for chaotic time series prediction. The proposed model employs an Online Interactive Recurrent Sequential learning method (OIRS-IT2FELM) to ensure rapid training with minimal parameter tuning and learning parameters. Specifically, the antecedent parameters of the type 2 membership functions in the BELIT2FELM model are initialized randomly, while the consequent type 1 parameters are derived analytically. The model's effectiveness is validated through experiments on two benchmark chaotic time series and a real-world application in electricity price forecasting for New York City. Results demonstrate that BELIT2FELM achieves lower root mean squared error (RMSE) compared to traditional BELT1FELM and other advanced models, particularly in the presence of noise.