Hybrid Intelligent System Combined with Perturbation Theory for Time Series Forecasting

• Paulo S. G. de Mattos Neto UFPE
• Aranildo Rodrigues L. J. UFRPE
• Tiago A. E. Ferreira UFRPE
• Germano C. Vasconcelos UFPE

Abstract

Most of the approaches described in the literature for time series prediction only employ the time series of interest to perform forecasting, discarding the error series coming from differences between the actual series and the model forecasting. Inspired on the Perturbation Theory, the Perturbative Time-lag Added Evolutionary Forecasting Method (P-TAEF Method) is a new method which applies treatment of the error series in conjunction with the original series to construct a more accurate time series forecasting model. Experiments conducted with the P-TAEF Method are compared to conventional artificial intelligence techniques illustrating the boosted perfomance of the method.

References

Battiti, R. (1992). “One step secant conjugate gradient”. Neural Computation, 4:141–166.

de Mattos Neto, P. S. G., Petry, G. G., Andrade, R. A., Ferreira, T. A. E., and Vasconcelos, G. C. (2007). “Hybrid swarm system for time series forecasting”. In Anais do VI Encontro Nacional de Inteligência Artificial - VI ENIA, Rio de Janeiro - RJ. Sociedade Brasileira de Redes Neurais.

Ferreira, T. A. E., Vasconcelos, G. C., and Adeodato, P. J. L. (2008). “A new intelligent system methodology for time series forecasting with artificial neural networks”. Neural Processing Letters, 28:113–129.

Hakykin, S. (2002). Redes Neurais - Princípos e Prática. Bookman, Porto Alegre - Brasil, 2a edition.

Hénon, M. (1976). “A two-dimensional mapping with a stranger attractor”. Communications in Mathematical Physics, 50(1):69–77.

Kantz, H. and Schreiber, T. (2003). Nonlinear Time Series analysis. Cambridge University Press, New York, NY, USA, second edition.

Leung, F. H. F., Lam, H. K., Ling, S. H., and Tam, P. K. S. (2003). “Tuning of the structure and parametrs of the neural network using an improved genetic algorithm”. IEEE Transaction on Neural Networks, 14(1):79–88.

Moller, M. F. (1993). “A scaled conjugate gradient algorithm for fast supervised learning”. Neural Networks, 6:525–533.

Mor, J. J. (1977). “The levenberg-marquardt algorithm: Implementation and theory”. In Watson, G. A., editor, Numerical Analysis, volume 630 of Lecture Notes in Mathematics, pages 105–116, Berlin. Springer-Verlag.

Reidmiller, M. and Braun, H. (1993). “A direct adaptive method for faster backpropagation learng: The rprop algorithm”. In Proceedings of the IEEE Int. Conf. on Neural Networks (ICNN), pages 586–591, San Francisco.

Sakurai, J. J. (1994). Modern Quantum Mechanics. Addison-Wesley, Massachusetts.

Savit, R. and Green, M. (1991). “Time series and dependent variables”. Physica D, 50:95–116.