A New Hybrid Model for Forecasting Crude Oil Price and the Techniques in the Model

Abstract:

Being able to predict crude oil prices with a reputation of intransigence to analysis or the directions of changing in crude oil price is of increasing value. We seek a method to forecast oil prices with precise predictions. In this paper, a hybrid model was proposed, which firstly decomposes the crude oil prices into several time series with different frequencies,then predict these time series which are not white noises, and at last integrate the predictions as the final results. We use Ensemble Empirical Mode Decomposition (EEMD) and Empirical Mode Decomposition (EMD) separately as the technique to decompose crude oil prices. Then we use Dynamic Artificial Neural Network (DAN2) and Back Propagation (BP) Neural Network separately as the technique to predict the deposed time series, and finally integrate the predictions produced by DAN2 or BP by Adaptive Linear Neural Network (ALNN) as the final result of predictions. EEMD has been proved as a very useful method to decompose the nonlinear and non-stationary time series, and DAN2, different from traditional artificial neural networks, also has obvious advantages over traditional ones. In this paper, EEMD and DAN2 are used to predict crude oil prices at the first time。 All in all, we build four models-EEMD-DAN2-ALNN, EMD-BP-ALNN, EEMD-BP-ALNN and EMD-DAN2-ALNN to test which technique, EMD or EEMD, could do better job in decomposition of crude oil prices in this kind of hybrid model and whetherDAN2 could outshine BP when used in this hybrid model. Experimental results of four hybrid models indicate EEMD-DAN2-ALNN could gives the most precise predictions of crude oil prices, and DAN2 has a better performance than traditional neural networks-BP,when used in this hybrid model and EEMD could do a better job than EMD in decomposition of crude oil prices to yield precise predictions of crude oil prices in this model.

• Periodical: Advanced Materials Research (Volume 974)
• Pages: 310-317
• DOI: https://doi.org/10.4028/www.scientific.net/AMR.974.310
• Online since: June 2014
• Authors: Jing Wen Zheng*, Shi Xiao Li, Yang Kun
• Keywords: Back Propagation Neural Network, Dynamic Artificial Neural Network, Empirical Mode Decomposition (EMD), Ensemble Empirical Mode Decomposition, Hybrid Model, Time Series

References

[1] Contreras, J., Espinola, R., Nogales, F. J., and Conejo, A. J: IEEE Transactions Vol. 18(2003), pp.1014-1020.

[2] Morana, C: Energy Economics Vol. 23(2001), p.325.

[3] Koutroumanidis, T., Ioannou, K., andArabatzis, G: Energy Policy Vol. 37(2009), p.3627.

[4] Mirmirani, S., and Li, H. C.: Advances in EconometricsVol. 19(2005), p.203.

[5] Yu, L., Wang, S., and Lai, K. K.: Forecasting crude oil price with an EMD-based neural network ensemble learning paradigm, " Energy Economics Vol. 30(2008), p.2623.

DOI: 10.1016/j.eneco.2008.05.003

[6] Zhang, X., Lai, K. K., and Wang, S. Y.: Energy economics Vol. 30(2008), p.905.

[7] Ghiassi, M., Saidane, H., andZimbra, D. K., A dynamic artificial neural network model for forecasting time series eventsm, International Journal of Forecasting, Vol. 21(2005), p.341.

DOI: 10.1016/j.ijforecast.2004.10.008

[8] Ghiassi, M., and Saidane, H.,: Neurocomputing, Vol. 63(2005), p.397.

[9] Huang, N. E., Shen, Z., Long, S. R., Wu, M. C., Shih, H. H., Zheng, Q., andLiu, H. H.,: Proceedings of the Royal Society of London Vol. 454(1998), p.903.

[10] Huang, N. E., Wu, M. L., Qu, W., Long, S. R., andShen, S. S.,: Applied Stochastic Models in Business and Industry, " Vol. 19(2003), pp.245-268.

[11] Wu, Z., and Huang, N. E.: Advances in Adaptive Data Analysis, " Vol. 1(2009), p.1.

[12] Ghiassi, M., Zimbra, D. K., andSaidane, H: Journal of Water Resources Planning and Management, Vol. 134(2008), p.138.

[13] Widrow, B: IEEE International Symposium on Circuits and Systems Vol. 2(1987).