Multi-step-ahead Prediction Based on Parallel Neural Network Ensemble
Ying Hua Qin · Computer Engineering and Applications Journal · 2006
By training a finite number of neural networks and then combining their results,neural network ensemble can significantly improve the generalization ability of learning systems.In this paper,a new model,Extraction of Characteristics Parallel Neural Network(ECPNN),is proposed for multi-step-ahead prediction.The framework of the model is composed of parallel Time Delay Neural Networks(TDNN) to process characteristic and remainder sequences extracted from single factor time series.Tested on time series of sunspot prediction,the model provides more accurate result for multi-step prediction than single TDNN.