Dendritic Neuron Model Trained by Biogeography-Based Optimization for Crude Oil Price Forecasting

Recent research reveals that single neuron with flexible dendritic plasticity can perform computation and information processing. Several dendritic neuron models have been proposed and achieving great success in various applications. All previous models use error back-propagation (BP) training method to adjust weights and thresholds in the model. Due to the inherent local search properties of BP, their performance usually suffers from the local optima problem. In this paper, we propose a biogeography-based optimization method to train the dendritic neuron model. Experiment is conducted for crude oil price forecasting and the results suggest that the proposed method can perform very well in comparison with the traditional multiple-layered perceptron.