Abstract
The accuracy of a model to forecast a time series diminishes as the prediction horizon increases, in particular when the prediction is carried out recursively. Such decay is faster when the model is built using data generated by highly dynamic or chaotic systems. This paper presents a topology and training scheme for a novel artificial neural network, named “Hybrid-connected Complex Neural Network” (HCNN), which is able to capture the dynamics embedded in chaotic time series and to predict long horizons of such series. HCNN is composed of small recurrent neural networks, inserted in a structure made of feed-forward and recurrent connections and trained in several stages using the algorithm back-propagation through time (BPTT). In experiments using a Mackey-Glass time series and an electrocardiogram (ECG) as training signals, HCNN was able to output stable chaotic signals, oscillating for periods as long as four times the size of the training signals. The largest local Lyapunov Exponent (LE) of predicted signals was positive (an evidence of chaos), and similar to the LE calculated over the training signals. The magnitudes of peaks in the ECG signal were not accurately predicted, but the predicted signal was similar to the ECG in the rest of its structure.
Similar content being viewed by others
References
Mees, AI (eds) (2000) Nonlinear dynamics and statistics. Birkhäuser, Boston
Galván IM, Isasi P (2001) Multi-step learning rule for recurrent neural models: an application to time series forecasting. Neural Process Lett 13: 115–133
Kinsner W (2006) Characterizing chaos through Lyapunov metrics. IEEE Trans Syst Man Cybernet C 36(2): 141–151
Strogatz SH (1994) Nonlinear dynamics and chaos: with applications to physics, biology, chemistry, and engineering. Addison-Wesley, New York
Principe JC, Kuo JM (1995) Dynamic modeling of chaotic time series with neural networks. In: Tesauro G, Touretzky DS, Leen TK (eds) Advances in neural information processing systems 7. Morgan Koufmann, San Mateo
Glass L, Mackey M (1988) From clocks to chaos. Princeton University, Princeton
Duhau S (2003) An early prediction of maximum sunspot number in solar cycle 24. Solar Phys 213: 203–212
Júnior MP-J, Barreto GA (2008) Long-term time series prediction with the NARX network: an empirical evaluation. Neurocomputing 71(16-18): 3335–3343
Abarbanel, Henry DI (2000) Challenges in modeling nonlinear systems: a worked example. In: Mees AI (ed) Nonlinear dynamics and statistics. Birkhauser, Boston
Judd K, Small M (2000) Towards long-term prediction. Physica D 136: 31–44
Granger CJW (2005) Preface: somethoughts on the future of forecasting. Oxford Bull Econ Stat 67(supplement(0305-9049)): 707–711
Mandic D, Chambers J (2001) Recurrent neural networks for prediction. Wiley, Chichester
Beliaev I, Kozma R (2007) Time series prediction using chaotic neural networks on the CATS benchmark. Neurocomputing 70: 2246–2439
Lendasse A, Oja E, Simula O, Verleysen M (2004) Time series prediction competition: the CATS Benchmark. In: Proceedings of IEEE IJCNN, pp 1615–1620
Leontaritis J, Billings SA (1985) Input–output parametric models for nonlinear systems—part I: deterministic nonlinear systems. Int J Control 41(2): 303–328
Principe JC, Euliano NR, Lefebvre WC (2000) Neural adaptive systems: fundamentals through simulations. Wiley, Chichester
Elman JL (1990) Finding structure in time. Cogn Sci 14: 179–211
Park DC, Tran CN, Lee Y (2006) Multiscale bilinear recurrent neural networks and their application to the long-term prediction of network traffic. In: LNCS 3973, pp 196–201
Cai X, Zhang N, Venayagamoorthy GK, Wunsch DC II (2004) Time series prediction with recurrent neural networks using a hybrid PSO-EA algorithm. In: Proceedings of IEEE IJCNN, pp 1647–1652. doi:10.1109/IJCNN.2004.1380208
Alarcon-Aquino V, Barria JA (2006) Multiresolution FIR neural-network-based learning algorithm applied to network traffic prediction. IEEE Trans Syst Man Cybernet C 36(2): 208–220. doi:10.1109/TSMCC.2004.843217
Sapankevych NI, Sankar R (2009) Time series prediction using support vector machines: a survey. IEEE Comput Intell Mag 4(2): 24–38
Gomez-Gil P (1998) The effect of non-linear dynamic invariant in the recurrent neural networks for prediction of electrocardiograms. Dissertation in Computer Science, Texas Tech University, Lubbock, TX, USA
Werbos P (1990) Backpropagation through time: what it does and how to do it. Proc IEEE 74(19): 1550–1560
Pearlmutter B (1990) Dynamic recurrent neural networks. Technical Report CMU-CS-90-196. School of Computer Science, Carnegie Mellon University, Pittsburgh, MA
Takens F (1981) Detecting strange attractors in turbulence. Springer, Berlin, pp 366–381
Abarbanel HDI, Brown R, Kadtke J (1990) Prediction in chaotic nonlinear systems: methods for time series with broad band Fourier spectra. Phys Rev A 41: 1782–1807
Parker TS, Chua LO (1989) Practical numerical algorithms for chaotic systems. Springer, New York
Gencay R, Dechert WD (1992) An algorithm for the n Lyapunov exponents of an n-dimensional unknown dynamical system. Physica D 59: 142–157
Wolf A, Swift JB, Swinney HL, Vastano JA (1985) Determining Lyapunov exponents from a time series. Physica D 16: 285–317
Sano J, Swada Y (1985) Measurement of the Lyapunov spectrum from a chaotic time series. Phys Rev Lett 55: 1082–1085
Rosestein M, Collins J, Deluca C (1993) A practical method for calculating largest Lyapunov exponents for small data set. Physica D 65: 117–134
Darbyshire AG, Broomhead DS (1996) Robust estimation of tangent maps and Lyapunov spectra. Physica D 89: 287–305
Gómez-Gil P, Oldham WB (1999) On the use of hybrid neural networks and non linear invariants for prediction of electrocardiograms. In: Proceedings of IJCNN 99, IEEE, pp 3661–3664. doi:10.1109/IJCNN.1999.836264
Logar A (1992) Recurrent neural networks and time series prediction. PhD Dissertation in Computer Science, Texas Tech University, Lubbock, TX
Werbos P (1994) The roots of backpropagation from ordered derivatives to neural networks and political forecasting. Wiley-Interscience Publication, New York
Sollacher R, Gao H (2008) Efficient online learning with spiral recurrent neural networks. In: IJCNN (2008), IEEE, pp 2551–2558. doi:10.1109/IJCNN.2008.4634155
García-Treviño ES, Alarcón-Aquino V (2006) Single-step prediction of chaotic time series using wavelet-networks. In: Electronics, robotics and automotive mechanics conference CERMA 2006 (Cuernavaca 2006), IEEE, pp 243–248
Glass L, Hunter P, McCulloch A (1991) Theory of heart. Springer, New York
Hongxuan Z, Yisheng Z, Yuhong X (2002) Complexity information based analysis of pathological ECG rhythm for ventricular tachycardia and ventricular fibrillation. Int J Bifurc Chaos 12(10): 2293–2303
Forberg JL, Green M, Bjork J, Ohlsson M, Edenbrandt L, Ohlin H, Ekelund U (2009) Search of the best method to predict acute coronary syndrome using only the electrocardiogram from the emergency department. J Electrocardiol 42(1): 58–63
Shanthi D, Sahoo G, Saravanan N (2009) Designing an artificial neural network model for the prediction of thrombo-embolic stroke. Int J Biom Bioinformatics 3(1): 10–18
Lerma C, Wassel N, Schirdewan A, Kurths J, Glass L (2008) Ventricular arrhythmias and changes in heart rate preceding ventricular tachycardia in patients with an implantable cardioverter defibrillator. Med Biol Eng Comput 46: 715–727. doi:10.1007/s11517-008-0326-y
Al-Khatib SM, Sanders GB, Bigger JT et al (2007) Preventing tomorrow’s sudden cardiac death today: part I: current data on risk stratification for sudden cardiac death. Am Heart J 153: 941–950
Harvard-MIT Division of Health Sciences Technology: (1992) The MIT_BIH Arrhythmia Database CD-ROM, 2nd edn. Biomedical Engineering Center, Cambridge, MA
García-Pedrero A, Gómez-Gil P (2010) Time series forecasting using recurrent neural network and wavelet reconstructed signals. In: Proceedings of 20th international conference on Electronics, communications and computers (Puebla, Mexico 2010), IEEE, pp 169–173. doi:10.1109/CONIELECOMP.2010.5440775
Smith LA (2000) Disentangling uncertainly and error: on the predictability of nonlinear systems. In: Mees AI (eds) Nonlinear dynamics and statics. Birkhäuser, Boston
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Gómez-Gil, P., Ramírez-Cortes, J.M., Pomares Hernández, S.E. et al. A Neural Network Scheme for Long-Term Forecasting of Chaotic Time Series. Neural Process Lett 33, 215–233 (2011). https://doi.org/10.1007/s11063-011-9174-0
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11063-011-9174-0