Problemas de investigación en la predicción de series de tiempo con redes neuronales artificiales

En el modelado de series de tiempo, las red es neuronales han venido ganando cada vez mas terreno, debido a sus reconocidas capacidades de adaptabilidad, generalizacion y aprendizaje. Si bien, en la literatura se denota un creciente interes por el desarrollo de aplicaciones con dichos modelos y se han presentado muchos reportes exitosos de su desempeno, igualmente, se han reportado resultados inconsistentes acerca de su uso. Algunos autores sostienen que tales inconsistencias son producto de falencias en la implementacion del modelo y en la carencia de una metodologia valida. En este trabajo se evaluan los efectos de dos factores clave en la construccion del modelo de red neuronal: el algoritmo de entrenamiento y el criterio de seleccion del numero de neuronas ocultas; para ejemplificar la discusion se uso la serie de pasajeros en lineas aereas de Box & Jenkins y los metodos de entrenamiento de regla delta generalizada y RProp. La evidencia experimental demuestra que los metodos de entrenamiento evaluados exhiben comportamientos diferentes a los teoricamente esper ados.

[1]  M. Rast.  Forecasting financial time series with fuzzy neural networks , 1997, 1997 IEEE International Conference on Intelligent Processing Systems (Cat. No.97TH8335).

[2]  Vassilis S. Kodogiannis,et al.  Forecasting Financial Time Series using Neural Network and Fuzzy System-based Techniques , 2002, Neural Computing & Applications.

[3]  Robert J. Marks,et al.  Electric load forecasting using an artificial neural network , 1991 .

[4]  David E. Rumelhart,et al.  Generalization by Weight-Elimination with Application to Forecasting , 1990, NIPS.

[5]  Amir F. Atiya,et al.  Multi-step-ahead prediction using dynamic recurrent neural networks , 2000, Neural Networks.

[6]  Venkat Subramanian,et al.  A GRG2-Based System for Training Neural Networks: Design and Computational Experience , 1993, INFORMS J. Comput..

[7]  Christian Lebiere,et al.  The Cascade-Correlation Learning Architecture , 1989, NIPS.

[8]  Martin A. Riedmiller,et al.  Advanced supervised learning in multi-layer perceptrons — From backpropagation to adaptive learning algorithms , 1994 .

[9]  Philippe Lauret,et al.  Bayesian neural network approach to short time load forecasting , 2008 .

[10]  Chris Chatfield,et al.  Time series forecasting with neural networks: a comparative study using the air line data , 2008 .

[11]  Michael Y. Hu,et al.  Forecasting with artificial neural networks: The state of the art , 1997 .

[12]  Nikolaos Kourentzes,et al.  Input-variable specification for Neural Networks - An analysis of forecasting low and high time series frequency , 2009, 2009 International Joint Conference on Neural Networks.

[13]  Roberto Battiti,et al.  First- and Second-Order Methods for Learning: Between Steepest Descent and Newton's Method , 1992, Neural Computation.

[14]  John E. Dennis,et al.  Numerical methods for unconstrained optimization and nonlinear equations , 1983, Prentice Hall series in computational mathematics.

[15]  S.K. Patra,et al.  Short term load forecasting using a novel recurrent neural network , 2008, TENCON 2008 - 2008 IEEE Region 10 Conference.

[16]  Shixin Cheng,et al.  Dynamic learning rate optimization of the backpropagation algorithm , 1995, IEEE Trans. Neural Networks.

[17]  David Zimbra,et al.  A dynamic artificial neural network model for forecasting time series events , 2005 .

[18]  Kurt Hornik,et al.  Multilayer feedforward networks are universal approximators , 1989, Neural Networks.

[19]  Kin Keung Lai,et al.  A neural-network-based nonlinear metamodeling approach to financial time series forecasting , 2009, Appl. Soft Comput..

[20]  Fred Collopy,et al.  How effective are neural networks at forecasting and prediction? A review and evaluation , 1998 .

[21]  Mikko Lehtokangas,et al.  Fast initialization for cascade-correlation learning , 1999, IEEE Trans. Neural Networks.

[22]  Marie Cottrell,et al.  Neural modeling for time series: A statistical stepwise method for weight elimination , 1995, IEEE Trans. Neural Networks.

[23]  Martin T. Hagan,et al.  Neural network design , 1995 .

[24]  William Remus,et al.  Neural Network Models for Time Series Forecasts , 1996 .

[25]  Milton S. Boyd,et al.  Designing a neural network for forecasting financial and economic time series , 1996, Neurocomputing.

[26]  Ramazan Gençay,et al.  Nonlinear modelling and prediction with feedforward and recurrent networks , 1997 .

[27]  JOHN G. CARNEY,et al.  Tuning Diversity in Bagged Ensembles , 2000, Int. J. Neural Syst..

[28]  Xuanxi Ning,et al.  A Framework for Time Series Forecasts , 2008, 2008 ISECS International Colloquium on Computing, Communication, Control, and Management.

[29]  Zhen Wang,et al.  Time series forecasting with RBF neural network , 2005, 2005 International Conference on Machine Learning and Cybernetics.

[30]  Emile Fiesler,et al.  Neural Network Initialization , 1995, IWANN.

[31]  Guoqiang Peter Zhang,et al.  An investigation of model selection criteria for neural network time series forecasting , 2001, Eur. J. Oper. Res..

[32]  T. Zhang,et al.  Forecasting time series by Bayesian neural networks , 2002, Proceedings of the 2002 International Joint Conference on Neural Networks. IJCNN'02 (Cat. No.02CH37290).