| Abstract | This paper presents the first in a series of experiments to study the behavior of a hybrid technique for model discovery in multivariate time series using similarity based neurofuzzy neural networks and genetic algorithms. This method discovers dependency patterns relating future values of a target series with past values of all examined series, and then constructs a prediction function. It accepts a mixture of numeric and non-numeric variables, fuzzy information, and missing values. Experiments were made changing parameters controlling the algorithm from the point of view of: i) the neuro-fuzzy network, ii) the genetic algorithm, and iii) the parallel implementation. Experimental results show that the method is fast, robust and effectively discovers relevant interdependencies. |
|---|
