DOI:
https://doi.org/10.14483/2322939X.4702Publicado:
2013-10-15Número:
Vol. 10 Núm. 1 (2013)Sección:
Actualidad TecnológicaPRONÓSTICO DE SERIES DE TIEMPO CON REDES NEURONALES REGULARIZADAS Y VALIDACIÓN CRUZADA
Palabras clave:
Pronóstico, Series de tiempo, regularización, validación cruzada, descomposición de pesos. (es).Descargas
Resumen (es)
En este trabajo se propone usar integralmente la estrategia de regularización de descomposición de pesos y validación cruzada con el fin de controlar integralmente el problema del sobreajuste en redes neuronales tipo perceptrón multicapa para el pronóstico de series de tiempo. Con el fin de evaluar la capacidad de la propuesta, se pronostica una serie de tiempo tradicional de la literatura. Los resultados evidencian que la combinación de ambas técnicas permite encontrar modelos con mejor capacidad de generalización que aproximaciones tradicionales.
Referencias
A. K. Palit y D. Popovic, Computational Intelligence in Time Series Forecasting, London: Springer, 2005.
D. C. Montgomery, L. A. Johnson y J. S. Gardiner, Forecasting & Time Series Analysis, Segunda ed., Singapure: Mc-Graw-Hill, Inc., 1990.
R. Gençay y T. Liu, «Nonlinear modeling and prediction with feedforward and recurrent networks,» Physica D: Nonlinear Phenomena, vol. 108, nº 1-2, pp. 119-134 , 1997.
B. L. Bowerman, R. T. O’Connell y A. B. Koehler, Forecasting, Time Series, and Regression: An Applied Approach, Cuarta ed., Ohio: Cengage Learning Brooks Cole, 2006.
N. Kasabov, Foundations of Neural Networks, Fuzzy Systems, and Knowledge Engineering, 2da ed., Massachusetts: The MIT Press Cambridge, 1998.
C. Granger y A. Anderson, An Introduction to Bilinear Time Series Models, Gottingen: Vandenhoeck and Ruprecht, 1978.
H. Tong y K. Lim, «Threshold autoregressive, limit cycles and cyclical data,» Journal of the Royal Statistical Society Series B, 42 (3). Pág. 245–292, vol. 42, nº 3, p. 245–292, 1980.
R. Engle, «Autoregressive conditional heteroskedasticity with estimates of the variance of UK inflation,» Econometrica, vol. 50, p. 987–1008, 1982.
K. S. Chan y H. Tong, «On estimating thresholds in autoregressive models,» Journal of Time Series Analysis, nº 7, pp. 178-190, 1986.
D. Dick van Dijk, T. Teräsvirta y F. Franses, «Smooth Transition Autoregressive Models - A Survey Of Recent Developments,» Econometric Reviews, nº 21, pp.1-47, 2002
H. Tong, «Threshold models in time series analysis —30 years on (with discussions by P.Whittle, M.Rosenblatt, B.E.Hansen, P.Brockwell, N.I.Samia & F.Battaglia),» Statistics & Its Interface, nº 4, pp. 107-136, 2011.
T. Bollerslev, «Generalised autoregressive conditional heteroscedasticity,» Journal of Econometrics, vol. 31, pp. 307-327, 1986.
H. Tong, Nonlinear Time Series: A Dynamical System Approach, Oxford: Oxford University Press, 1990.
I. De Gooijer y K. Kumar, «Some Recent Developments in Non- Linear Modelling, Testing, and Forecasting,» International Journal of Forecasting, vol. 8, pp. 135-156, 1992.
D. Peña, «Second-generation time-series models: a comment on ‘Some advances in non-linear and adaptive modelling in time-series analysis’ by Tiao and Tsay,» Journal of Forecasting, vol. 13, pp. 133-140, 1994.
D. Tjostheim, «Nonlinear time series: a selective review,» Scand. J. Statist., vol. 21, pp. 97-130, 1994.
W. Hardle, H. Liitkepohl y R. Chen, «A review of non-parametric time series analysis.,» Int. Statist. Rev., vol. 65, pp. 49-72, 1997.
C. Granger y T. Teräsvirta, Modelling Nonlinear Economic Relationships, Oxford: Oxford University Press, 1993.
C. Granger, «Strategies for modeling nonlinear time-series relationships,» The Economic Record, vol. 69, nº 206, p. 233–238, 1993.
P. Zhang, B. Patuwo y M. Hu, «A simulation study of artificial neural networks for nonlinear time-series forecasting,» Computers & Operations Research, vol. 28, nº 4, pp. 381-396, 2001.
M. Ghiassi, H. Saidane y D. Zimbra, «A dynamic artificial neural network model for forecasting time series events,» International Journal of Forecasting, vol. 21, nº 2, pp. 341-362, 2005.
F. A. Villa, J. D. Velásquez y R. C. Souza, «Una aproximación a la regularización de redes cascada-correlación para la predicción de series de tiempo.,» Investigación Operacional., nº 28, pp. 151-161, 2008.
T. Masters, Practical neural network recipes in C++, New York: Academic Press, 1993.
G. Zhang, B. E. Patuwo y M. Y. Hu, «Forecasting with artificial neural networks: the state of the art,» International Journal of Forecasting, vol. 14, nº 1, pp. 35-62, Marzo 1998.
S. Haykin, Neural Networks: A Comprehensive Foundation, New Jersey: Prentice Hall, 1999.
D. Zhang, Y. Han, X. Ning y .. Liu, «A Framework for Time Series Forecasts,» Proceedings ISECS International Colloquium on Computing, Communication, Control, and Management, vol. 1, pp. 52- 56, 2008.
X.-B. Yan, Z. Wang, S.-H. Yu y Y.-J. Li, «Time Series Forecasting with RBF Neural Network,» Proceedings of the Fourth International Conference on Machine Learning and Cybernetics, pp. 4680-4683, 2005.
M. Rast, «Forecasting Financial Time Series with Fuzzy Neural Networks,» IEEE lnternational Conference on Intelligent Processing Systems, pp. 432-434, 1997.
V. Kadogiannis y A. Lolis, «Forecasting financial time series using neural network and fuzzy system-based techniques, » Neural Computing and Application, vol. 11, pp. 90-102, 2002.
A. Parlos, O. Rais y A. Atiya, «Multistep-ahead prediction using dynamic recurrent neural networks,» Neural Networks, vol. 13, pp. 765-786, 2000.
S. Mishra y S. Patra, «Short term load forecasting using a novel recurrent neural network,» International Journal of Computational Intelligence: Theory and Practice, vol. 4, nº 1, pp. 39-45, 2009.
P. Sánchez y J. D. Velásquez, «Problemas de Investigación en la Predicción de Series de Tiempo con Redes Neuronales Artificiales,» Revista Avances en Sistemas e Informática, vol. 7, nº 3, pp. 67-73, 2010.
A. Tikhonov, «On Solving Incorrectly Posed Problems and Method of Regularization, » Docklady Academii Nauk, vol. 151, pp. 501-504, 1963.
G. Hinton, «Connectionist learning procedures, » Artificial Intelligence, nº 40, p. 185–243, 1989.
E. W. Steyeberg, Clinical Prediction Models, New York: Springer, 2009.
S. Geisser, Predictive Inference, New York, NY: Chapman and Hall, 1993.
R. Kohavi, «A study of cross-validation and bootstrap for accuracy estimation and model selection,» Proceedings of the Fourteenth International Joint Conference on Artificial Intelligence, vol. 2, nº 12, p.1137–1143, 1995.
P. A. Devijver y J. and Kittler, Pattern Recognition: A Statistical Approach, London, GB: Prentice-Hall, 1982.
F. A. Villa y J. D. Velásquez, «Regularización de Redes Cascada Correlación con Regresión en Cadena,» Quinto Congreso Colombiano de Computación (Abril 14 – 16), Cartagena, 2010.
K. Hornik, M. Stinchcombe y H. White, «Multilayer feedforward networks are universal approximators,» Neural Networks, vol. 2, p. 359–366, 1989.
G. Cybenko, «Approximation by superpositions of a sigmoidal function,» Mathematics of Control: Signals and Systems, vol. 2, p. 202–314, 1989.
K. Funahashi, «On the approximate realization of continuous mappings by neural networks,» Neural Neworks, vol. 2, p. 183–192, 1989.
W. Sarle, «The 19th Annual SAS Users Group Int. Conference,» de Neural networks and statistical models, Cary, North Carolina, 1994.
T. W. S. Chow y S. Cho, Neural networks and computing: learning algorithms and applications, vol. 7, London: Imperial College Press, 2007, p. 309.
M. Riedmiller y H. Braun, «A direct adaptive method for faster backpropagation learning: The RPROP algorithm, » de Proceedings of the IEEE International Conference on Neural Networks, IEEE Press, 1993, p. 586–591
M. Riedmiller, «Advanced supervised learning in multi-layer perceptrons – from backpropagation to adaptive learning algorithms,» Computer Standards and Interfaces, vol. 16 , p. 265–278, 1994.
D. M. Ortíz, F. A. Villa y J. D. Velásquez, «Una Comparación entre Estrategias Evolutivas y RPROP para la Estimación de Redes Neuronales,» Avances en Sistemas e Informática, vol. 4, nº 2, p. 135–144, 2007}
I. Kaastra y M. Boyd, «Designing a neural network for forecasting financial and economic series,» Neurocomputing, nº 10, pp. 215-236, 1996.
U. Anders y O. Korn, «Model selection in neural networks,» Neural Networks, nº 12, pp. 309-323, 1999.
V. Cherkassky y Y. Ma, «Another look at statistical learning theory and regularization, » Neural Networks, vol. 7, nº 22, pp. 958-969, 2009.
C. Leung, H. Wang y J. Sum, «On the selection of weight decay parameter for faulty networks,» IEEE Transactions on Neural Networks , vol. 8, nº 21, pp. 1232- 1244 , 2010.
C. M. Bishop, Neural Networks for Pattern Recognition, New York: Oxford University Press Inc., 1994, p. 482.
A. E. Hoerl y R. W. Kennard, «Ridge Regression: Biased Estimation for Nonorthogonal Problems,» Technometrics., vol. 12, nº 1, p. 55–67, 1970.
G. Karystinos y D. Pados, «On overfitting, generalization, and randomly expanded training sets,» IEEE Transactions on Neural Networks, vol. 11, nº 5, pp. 1050-1057, 2000.
M. J. Campbell y A. M. Walker, «A survey of statistical work on the Mackenzie river series of annual canadian lynx trappings for the years 1821–1934 and a new analysis,» Journal of the Royal Statistical Society, vol. 140, nº 4, p. 411–431, 1977.
T. S. Rao y M. Gabr, «An introduction to bispectral analysis and bilinear time series models,» Lecture Notes in Statistics, vol. 24, p. 528–535, 1984.
G. Zhang, «Time Series forecasting using a hybrid ARIMA and neural network model,» Neurocomputing, vol. 50, p. 159–175, 2003.
2.png)
