O uso de algoritmos genéticos para a solução de problemas de agrupamento automático

Abstract

The Clustering consists to organize a set of information into groups according to the similarities present in their records, so that common properties between the data can be identified for easy understanding. Sometimes the number of groups is a unavailable data for the resolution of this process. When this information is unknown a priori we have the so-called Automatic Clustering Problem. In this work are presented algorithms to solve this problem, based on the meta-heuristic Genetic Algorithm. The quality of a Clustering can be affected by the way the initial groups are generated and by the choice of the cluster validity index, different methods are proposed based on an already existing methodology with the purpose of generating better quality resolutions. Experiments were applied to proposed approaches for several data sets. The obtained results are compared with each other and with other works of the literature.