Adaptive neural network model for time-series forecasting

• Published in: European journal of operational research Vol. 207; no. 2; pp. 807 - 816
• Main Authors: Wong, W.K., Xia, Min, Chu, W.C.
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 01.12.2010; Elsevier; Elsevier Sequoia S.A
• Series: European Journal of Operational Research
• Subjects: Adaptive metrics; Admixtures; Amplitudes; Applied sciences; Artificial neural networks; Exact sciences and technology; Forecasting; Inference from stochastic processes; time series analysis; Learning theory; Mathematical models; Mathematics; Neural networks; Operational research; Operational research. Management science; Operations research; Probability and statistics; Regression analysis; Robustness; Sciences and techniques of general use; Statistics; Studies; Time series; Time-series Forecasting Adaptive metrics Neural networks; Time-series; Neural networks; Forecasting; Adaptive metrics; Chaos; Nearest neighbour; Periodical; Time series; Autoregressive model; Regression analysis; Neural network; Real time; Modeling; Adaptive method; Relative error; Metric; Robustness
• ISSN: 0377-2217; 1872-6860
• DOI: 10.1016/j.ejor.2010.05.022

In this study, a novel adaptive neural network ( ADNN) with the adaptive metrics of inputs and a new mechanism for admixture of outputs is proposed for time-series prediction. The adaptive metrics of inputs can solve the problems of amplitude changing and trend determination, and avoid the over-fitting of networks. The new mechanism for admixture of outputs can adjust forecasting results by the relative error and make them more accurate. The proposed ADNN method can predict periodical time-series with a complicated structure. The experimental results show that the proposed model outperforms the auto-regression ( AR), artificial neural network ( ANN), and adaptive k- nearest neighbors ( AKN) models. The ADNN model is proved to benefit from the merits of the ANN and the AKN through its’ novel structure with high robustness particularly for both chaotic and real time-series predictions.