Abstract
A key problem in time series prediction using autoregressive models is to fix the model order, namely the number of past samples required to model the time series adequately. The estimation of the model order using cross-validation may be a long process. In this paper, we investigate alternative methods to cross-validation, based on nonlinear dynamics methods, namely Grassberger–Procaccia, Kégl, Levina–Bickel and False Nearest Neighbors algorithms. The experiments have been performed in two different ways. In the first case, the model order has been used to carry out the prediction, performed by a SVM for regression on three real data time series showing that nonlinear dynamics methods have performances very close to the cross-validation ones. In the second case, we have tested the accuracy of nonlinear dynamics methods in predicting the known model order of synthetic time series. In this case, most of the methods have yielded a correct estimate and when the estimate was not correct, the value was very close to the real one.
Similar content being viewed by others
Notes
A manifold (http://en.wikipedia.org/wiki/Manifold) is a mathematical space in which every point has a neighborhood which resembles Euclidean space, but in which the global structure may be more complicated. In a one-dimensional manifold (e.g. a line, a circle) every point has a neighborhood that looks like a segment of a line. In a two-dimensional manifold (e.g. a plane, the surface of a sphere) the neighborhood looks like a disk. \({\mathbb{R}}^n\) is a n-dimensional manifold.
M is diffeomorphic to U iff there is a differentiable map \(m : M\,\mapsto\,U\) whose inverse m −1 exists and is also differentiable.
Takens–Mañé embedding theorem is a consequence of a Whitney Embedding Theorem [32] stating that a generic map from an S-dimensional manifold to a (2S + 1)-dimensional Euclidean space is an embedding, i.e. the image of the S-dimensional manifold is completely unfolded in the larger space. Therefore, two points in the S-dimensional manifold do not map to the same point in the (2S + 1)-dimensional space.
I(λ) is 1 iff condition λ holds, 0 otherwise.
The complexity of effective sorting algorithms (e.g. mergesort and heapsort) is \(\ell \log \ell\), where \(\ell\) is the number of elements that have to be sorted.
In our implementation, we use the linear search having complexity \(O(d \ell^2).\)
Paris-14E Parc Montsouris and DSVC1 time series can be downloaded from http://www.knmi.nl/samenw/eca and http://www.cpdee.ufmg.br/∼MACSIN/services/data/data.htm, respectively.
Windows Vista is a registered trademark of Microsoft Inc.
Mathematica is a registered trademark of Wolfram Inc.
Available on request for further investigations and comparisons.
References
Aguirre LA, Rodrigues GG, Mendes EM (1997) Nonlinear identification and cluster analysis of chaotic attractors from a real implementation of Chua’s circuit. Int J Bifurcat Chaos 6(7):1411–1423
Akaike H (1974) A new look at the statistical model identification. IEEE Trans Autom Control 19(6):716–723
Anderson TW (1963) Determination of the order of dependence in normally distributed time series. In: Rosenblatt M (ed) Time series Analysis. Wiley, London, pp 425–446
Arbabanel HDI (1996) Analysis of observed chaotic data. Springer, Berlin
Chua LO, Komuro M, Matsumoto T (1985) The double scroll. IEEE Trans Circuits Syst 32(8):797–818
Duda RO, Hart PE, Stork DG (2000) Pattern Classification. Wiley, New York
Eckmann JP, Ruelle D (1985) Ergodic theory of Chaos and strange attractors. Rev Mod Phys 57:617–659
Grassberger P, Procaccia I (1983) Measuring the strangeness of strange attractors. Physica D9:189–208
Hirshberg D, Merhav N (1996) Robust methods for model order estimation. IEEE Trans Signal Proces 44:620–628
Hübner U, Weiss CO, Abraham NB, Tang D (1994) Lorenz-Like Chaos in NH3-FIR Lasers. Time Series Prediction. Forecasting the Future and Understanding the Past. Addison Wesley, Reading, pp 73–104
Joachim T (1999) Making large-scale SVM learning practical. Advances in Kernel methods—support vector learning, MIT Press, Cambridge
Kantz H, Schreiber T (1997) Nonlinear time series analysis. Cambridge University Press, London
Kennel MB, Brown R, Arbabanel HDI (1992) Determining embedding dimension for phase-space reconstruction using a geometrical construction. Phys Rev A 45(6):3403–3411
Kégl B (2003) Intrinsic dimension estimation using packing numbers, vol 15. Advances in Neural Information Processing. MIT Press, Cambridge
Mañé R (1981) On the dimension of compact invariant sets of certain nonlinear maps. Dynamical Systems and Turbolence, Warwick 1980. In: Lecture Notes in Mathematics no. 898, Springer, Berlin, pp 230–242
Merhav N (1989) The estimation of the model order in exponential families. IEEE Trans Inf Theory 35(5):1109–1114
Müller K-R, Rätsch G, Kohlmorgen J, Smola A, Schölkopf B, Vapnik V (2000) Time series prediction using support vector regression and neural network. In: Higuchi T, Takizawa Y (eds) Proceedings of second international symposium on frontiers of time series modelling: nonparametric approach to knowledge discovery. Institute of mathematical statistic publication
Ott E (1993) Chaos in dynamical systems. Cambridge University Press, London
Packard N, Crutchfield J, Farmer J, Shaw R (1980) Geometry from a time series. Phys Rev Letters 45(1):712–716
Schölkopf B, Smola A (2002) Learning with Kernels. MIT Press, Cambridge
Rissanen J (1978) Modeling by shortest data description. Automatica 14(5):465–471
Shawe-Taylor J, Cristianini N (2004) Kernel Methods for Pattern Analysis. Cambridge University Press, London
Schwarz G (1978) Estimating the dimension of a model. Ann Stat 6(2):461–464
Smith RL (1992) Optimal estimation of fractal dimension. Nonlinear modeling and forecasting. In: Casdagli M, Eubank S (eds) SFI studies in the sciences of complexity, vol. XII, Addison-Wesley, New York, pp 115–135
Snyder DL (1975) Random point processes. Wiley, New York
Stone M (1974) Cross-validatory choice and assessment of statistical prediction. J R Stat Soc 36(1):111–147
Takens F (1981) Detecting strange attractor in turbolence. Dynamical systems and turbolence, Warwick 1980. In: Lecture Notes in Mathematics no. 898. Springer, Berlin, pp 366–381
Takens F (1985) On the numerical determination of the dimension of an attractor. Dynamical systems and bifurcations. In: Braaksma B, Broer H, Takens F (eds) Proceedings Groningen 1984, Lecture Notes in Mathematics No. 1125, Springer, Berlin, pp 99–106
Tong H (1990) Nonlinear time series. Oxford University Press, NY
Vapnik VN (1998) Statistical Learning Theory. Wiley, New York
Vaidya PM (1989) An O(n log n) algorithm for the All-Nearest-Neighbors Problem. Discret Comput Geom 4(1):101–115
Whitney H (1936) Differentiable manifolds. Ann Math 37:645–680
Wijngaard JB, Klein Tank AMG, Konnen GP (2003) Homogeneity of 20th century European daily temperature and precipitation series. Int J Clim 23:679–692
Acknowledgments
The authors wish to thank the anonymous reviewers for their valuable comments.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Camastra, F., Filippone, M. A comparative evaluation of nonlinear dynamics methods for time series prediction. Neural Comput & Applic 18, 1021–1029 (2009). https://doi.org/10.1007/s00521-009-0266-y
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00521-009-0266-y