Automated Linear Modeling of Time Series with Self Adaptive Genetic Algorithms

Abstract

Abstract-In this paper we present two algorithms that automatically calculate linear expressions for Time Series. To estimate the maximum number of terms of the linear expression and the intervals in which the series coefficients vary, the algorithms are based in the Box-Jenkins methodology. With this information and establishing beforehand the number of terms that are required, the Self Adaptive Genetic Algorithms are applied in several stages to obtain the series model. The proposed algorithms were tested in the Box-Jenkins classical examples, obtaining satisfactory results. It is worth it to mention that these algorithms allow treating series with time-dependent trends and variances. The methodology based on Self Adaptive Genetic Algorithms is used to estimate linear models for every example of NN3 2007, although in this paper we are presenting only the results of NN3-REDUCED.