Correo Electrónico
DNINFOA - SIA
Bibliotecas
Convocatorias
Identidad U.N.
Publicado
Modelado de cambios de nivel en series de tiempo no lineales usando redes neuronales artificiales
Palabras clave:
Series de tiempo, Cambios estructurales, Cambios en Nivel, Redes Neuronales Artificiales, Perceptrónes Multicapa, Series de tiempo-No lineales (es)Time Series, Structural Breaks, Level Shifts, Nonlinear Time Series, Artificial Neural Networks, Multilayer Perceptrons (en)
Descargas
Es comunmente aceptado que muchas variables físicas y económicas presentan comportamientos dinámicos no lineales, cuya complejidad hacen que sea imposible formular un modelo econométrico basado en leyes físicas o económicas que representen su evolución adecuadamente. El problema de la formulación del modelo se ve agravado por la presencia de observaciones atípicas y cambios estructurales, para las cuales no existen modelos matemáticos que permitan su representación en el caso no lineal. En este estudio, se presentan tres nuevos modelos que permiten la representación de cambios de nivel en series de tiempo no lineales usando perceptrónes multicapa, los cuales son ejemplificados para una serie ficticia y una serie real ampliamente conocida y estudiada. Los resultados indican las bondades de los modelos propuestos, permitiendo que ellos sean recomendados como parte integral de las metodologías para la modelación de series de tiempo usando redes neuronales artificiales.
It is usually accepted that many physical and economical variables shown nonlinear dynamical behaviors which complexity makes it impossible to formulate an econometric model based only on physical or economic laws that properly represents its evolution. The formulation of model is aggravated with the presence of outliers and structural breaks, for which there are not any mathematical models that allow its representation in the nonlinear case. In this paper, three new models are presented which allow the representation of level shifts in nonlinear time series using multilayer perceptrons, which are illustrated for one artificial and one real series, and it is proposed a methodological strategy for the construction and specification of such models, based on a previous linear modeling. The results indicate the goodness of the proposed models, allowing them to be recommended as an integral part of the methodologies for modeling time series using artificial neural networks.
Referencias
Box, G. y Jenkins, G. (1970), Time Series Analysis: Forecasting and Control, Holden-Day Inc, San Francisco.
Box, G. y Tiao, G. (1975), ‘Intervention analysis with applications to economic and environmental problems’, Journal of the American Statistical Association 70, 70-70.
Chen, C. y Liu, L. (1993), ‘Joint estimation and outlier effects in time series’, Journal of the American Statistical Association 88, 284-297.
Fox, A. (1972), ‘Outliers in time series’, Journal of the Royal Statistical Society Series B 34, 350-363.
Franses, P. y Van Dijk, D. (2000), Non-linear time series models in empirical finance, Cambridge University Press.
Granger, C. y Terasvirta, T. (1993), Modeling Nonlinear Economic Relationships, Oxford University Press.
Harvey, A. (1989), Forecasting, Structural Time Series and the Kalman Filter, Cambridge University Press.
Junttila, J. (2001), ‘Structural breaks, ARIMA model and finnish inflation forecasts’, International Journal of Forecasting 17, 203-230.
Kaastra, I. y Boyd, M. (1996), ‘Designing a neural network for forecasting financial and economic series’, Neurocomputing 10, 215-236.
Masters, T. (1995), Neural, Novel and Hybrid Algorithms for Time Series Prediction, John Wiley and Sons Inc.
Pole, A., West, M. y Harrison, J. (1994), Applied Bayesian Forecasting and Time Series Analysis, Chapman & Hall, NY.
Rodríguez, G. (2002), Modelos estructurales en el contexto de las series temporales económicas, Documento de trabajo, Universidad de La Laguna y Universidad de Las Palmas de Gran Canaria.
Tolvi, J. (1998), Outliers en time series: A review, Research reports 76, University of Turku, Department of Economic.
Tolvi, J. (2001), Nonlinear model selection in the presence of outliers, Research reports 90, University of Turku, Department of Economic.
Tong, II. (1990), Non-linear Time Series: A Dynamical System Approach, Oxford Statistical Science Series, Claredon Press Oxford.
Tsay, R. (1988), ‘Outliers, level shifts, and variance changes in time series', Journal of Forecasting 7, 1-20.
Vaage, K. (2000), ‘Detection of outliers and level shifts in time series: An evaluation of two alternative procedures', Journal of Forecasting 19, 23-37.
Van Dijk, D., Franses, P. y Lucas, A. (1990), Testing for smooth transition nonlinearity in the presence of outliers, Research reports Report 9622/A, Erasmus University, Rotterdam. Econometric Institute.
