The genetic algorithms have seen applied in knowledge discovery and specially for discovering association rules. In this paper, we explore the use of di erent rule quality measures in the tness function in a genetic algorithm for discovering association rules. Also, we present an improvement for this algorithm: (i) the mutation stage is calculated with a probability independent for each individual and (ii) the selection stage is calculated with Boltzmann selection. The proposed version was tested with 10 di erent rule quality evaluation functions on 6 benchmark datasets.
Eje: Workshop Bases de datos y minería de datos (WBDDM)
Red de Universidades con Carreras en Informática (RedUNCI)