Interval emotional neural network for prediction of Kp, AE and Dst geomagnetic activity indices

Abstract

In this paper, an interval emotional neural network (ENN) is proposed for uncertainty modeling of chaotic time series. The proposed interval ENN consist of 4 main neural blocks including: thalamus (TH), amygdala (AMY), orbitofrontal cortex (OFC) and sensory cortex (SC). In the proposed interval ENN, the main connections located between these components are modeled by interval weights including: TH-AMY, OFC-AMY, SC-OFC and SC-AMY connections. In this paper genetic algorithm (GA) is applied for optimal tuning of interval ENN's weights and a traditional artificial neural network (ANN) is used for comparison purposes. In the Experimental Results Section, chaotic geomagnetic activity Kp, AE and Dst indices are used as a case study. Numerical results indicate the superiority of the proposed interval ENN in term of higher accuracy and lower uncertainty.