Abstract
Predicting the future of a time-series is important in several and diverse applications. Whilst traditional methods are based on fixed linear models, techniques that use artificial neural networks are typically trained offline. We propose an online technique for time-series prediction using a single-layer feedforward neural network trained as an extreme learning machine, where the weights are updated with each new observation. The inputs to this machine not only include the actual observations of the time-series but as well as the predicted values from the machine. This ensures that feedback is incorporated into the training process. We demonstrate that such a procedure improves the prediction accuracy. Moreover, using multiple networks trained in this manner further improves the performance.
This is a preview of subscription content, log in via an institution to check access.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Box, G.E.P., Jenkins, G.M., Reinsel, G.C.: Time Series Analysis: Forecasting and Control. Prentice-Hall, Upper Saddle River (1994)
Connor, J.T., Martin, D.R., Atlas, L.E.: Recurrent neural networks and robust time series prediction. IEEE Transactions on Neural Networks 5, 240–254 (1994)
Farmer, J.D., Sidorowich, J.J.: Exploiting chaos to predict the future and reduce noise. In: Lee, Y.C. (ed.) Evolution, Learning, and Cognition, pp. 277–330. World Scientific, Singapore (1988)
George, K.: Some applications of multiple models methodology. In: Proceedings of the Fifteenth Yale Workshop on Adaptive and Learning Systems, pp. 81–86. Yale University, New Haven (June 2011)
Gershenfeld, N.A., Weigend, A.S.: The future of time series. In: Weigend, A.S., Gershenfeld, N.A. (eds.) Time Series Prediction: Forecasting the Future and Understanding the Past, pp. 1–70. Addison-Wesley (1993)
Hashem, S., Schmeiser, B.: Improving model accuracy using optimal linear combinations of trained neural networks. IEE Transactions on Neural Networks 6, 792–794 (1995)
Hornik, K., Stinchcombe, M., White, H.: Multilayer feedforward networks are universal approximators. Neural Networks 2, 359–366 (1989)
Huang, G.B., Zhu, Q.Y., Siew, C.K.: Extreme learning machine: A new learning scheme of feedforward neural networks. In: Proceedings of IEEE International Joint Conference on Neural Networks, pp. 985–990 (2004)
Liang, N.Y., Huang, G.B., Saratchandran, P., Sundararajan, N.: A fast and accurate online sequential learning algorithm for feedforward networks. IEEE Transactions on Neural Networks 17, 1411–1423 (2006)
Mackey, M.C., Glass, L.: Oscillation and chaos in physiological control systems. Science 197, 287–289 (1977)
Narendra, K.S., Balakrishnan, J.: Performance improvement in adaptive control systems using multiple models and switching. In: Proceedings of the Seventh Yale Workshop on Adaptive Learning Systems, pp. 27–33. Center for Systems Science, Yale University, New Haven, USA (May 1992)
Narendra, K.S., Driollet, O.A., Feiler, M., George, K.: Adaptive control of time-varying systems using multiple models. International Journal of Adaptive Control and Signal Processing 17(2), 87–102 (2003)
Narendra, K.S., Parthasarathy, K.: Identification and control of dynamical systems using neural networks. IEEE Transactions on Neural Networks 1(1), 4–27 (1990)
Shabri, A.: Comparison of time series forecasting methods using neural networks and Box-Jenkins model. Matematika UTM 17, 1–6 (2001)
Tang, Z., Fishwick, P.A.: Feed-forward neural nets as models for time series forecasting. ORSA Journal of Computing 5, 374–385 (1993)
Tong, H., Lim, K.S.: Threshold autoregression, limit cycles, and cyclical data. Journal of the Royal Statistical Society, B 42, 245–292 (1980)
Zhang, P.G., Patuwo, E.B., Hu, M.Y.: A simulation study of artificial neural networks for nonlinear time-series forecasting. Computers & Operations Research 28, 381–396 (2001)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2015 Springer International Publishing Switzerland
About this paper
Cite this paper
George, K., Prabhu, S., Mutalik, P. (2015). An Online Multiple-Model Approach to Univariate Time-Series Prediction. In: Cao, J., Mao, K., Cambria, E., Man, Z., Toh, KA. (eds) Proceedings of ELM-2014 Volume 1. Proceedings in Adaptation, Learning and Optimization, vol 3. Springer, Cham. https://doi.org/10.1007/978-3-319-14063-6_19
Download citation
DOI: https://doi.org/10.1007/978-3-319-14063-6_19
Publisher Name: Springer, Cham
Print ISBN: 978-3-319-14062-9
Online ISBN: 978-3-319-14063-6
eBook Packages: EngineeringEngineering (R0)