Forecasting Time Series by Means of Evolutionary Algorithms

The time series forecast is a very complex problem, consisting in predicting the behaviour of a data series with only the information of the previous sequence. There is many physical and artificial phenomenon that can be described by time series. The prediction of such phenomenon could be very complex. For instance, in the case of tide forecast, unusually high tides, or sea surges, result from a combination of chaotic climatic elements in conjunction with the more normal, periodic, tidal systems associated with a particular area. Too much variables influence the behaviour of the water level. Our problem is not only to find prediction rules, we also need to discard the noise and select the representative data. Our objective is to generate a set of prediction rules. There are many methods tying to achieve good predictions. In most of the cases this methods look for general rules that are able to predict the whole series. The problem is that usually the time series has local behaviours that don’t allow a good level of prediction when using general rules. In this work we present a method for finding local rules able to predict only some zones of the series but achieving better level prediction. This method is based on the evolution of set of rules genetically codified, and following the Michigan approach. For evaluating the proposal, two different domains have been used: an artificial domain widely use in the bibliography (Mackey-Glass series) and a time series corresponding to a natural phenomenon, the water level in Venice Lagoon.

[1]  D.E. Goldberg,et al.  Classifier Systems and Genetic Algorithms , 1989, Artif. Intell..

[2]  Inés María Galván,et al.  A Selective Learning Method to Improve the Generalization of Multilayer Feedforward Neural Networks , 2001, Int. J. Neural Syst..

[3]  Hans-Paul Schwefely,et al.  Evolutionary Algorithms: Some Very Old Strategies for Optimization and Adaptation , 1992 .

[4]  Norman H. Packard,et al.  A Genetic Learning Algorithm for the Analysis of Complex Data , 1990, Complex Syst..

[5]  Stephen F. Smith,et al.  A learning system based on genetic adaptive algorithms , 1980 .

[6]  David B. Fogel,et al.  An introduction to simulated evolutionary optimization , 1994, IEEE Trans. Neural Networks.

[7]  Cezary Z. Janikow,et al.  A knowledge-intensive genetic algorithm for supervised learning , 1993, Machine Learning.

[8]  Melanie Mitchell,et al.  An introduction to genetic algorithms , 1996 .

[9]  Kenneth A. De Jong,et al.  Using genetic algorithms for concept learning , 1993, Machine Learning.

[10]  John C. Platt.  A Resource-Allocating Network for Function Interpolation , 1991, Neural Computation.

[11]  K. De Jong,et al.  Using Genetic Algorithms for Concept Learning , 2004, Machine Learning.

[12]  Y Lu,et al.  A Sequential Learning Scheme for Function Approximation Using Minimal Radial Basis Function Neural Networks , 1997, Neural Computation.

[13]  Fernanda Strozzi,et al.  Forecasting high waters at Venice Lagoon using chaotic time series analysis and nonlinear neural networks , 2000 .

[14]  John H. Holland,et al.  Adaptation in Natural and Artificial Systems: An Introductory Analysis with Applications to Biology, Control, and Artificial Intelligence , 1992 .