Predicting chaotic time series by ensemble self generating neural networks merged with genetic algorithm

Self-generating neural networks (SGNNs) are focused attention because of their simplicity on networks design. Due to its instability, the ensemble networks are used to improve the prediction accuracy. In this paper, we analyzed the correlation between the ensemble components, then propose a method based on genetic algorithm to optimally merge the ensemble components. The experiments on two time series generated from Henon mapping, Ikeda mapping prove that the method effectively improves the prediction accuracy of time series.