Knowledge extraction in signals classification with genetic algorithms

In the analysis of signals from massive databases it is desirable to have automatic mechanisms for classification. The synergy of diverse artificial intelligence techniques with advanced signal representation models is becoming very efficient in developing this kind of task. In this paper, it is shown that genetic algorithms focused on rule discovery might be used for this purpose. In our approach, each individual represents a classifying rule, composed of an antecedent and a consequence. Using a technique based on niches in order to avoid the extinction of any of the species, we obtain several solutions that form an expert classification system. The results have been compared with those of other classifiers on the same signals and they show efficiency of our procedure.