Evolutionary support vector machines for diabetes mellitus diagnosis

Abstract. The paper proposes a first attempt to approach regression from the perspective of the novel technique of evolutionary support vec-tor machines (ESVMs). ESVMs have been developed with the aim of providing a simpler architecture as compared to their canonical coun-terpart. As a consequence, they inherit the training mechanism of SVMs but approach the resulting optimization problem through evolutionary algorithms. In this paper, ESVMs for proposed task are built upon the classical learning engine of -SVMs for regression. The validation of the new approach on a set of 2-dimensional points once again demon-strates the advantages of ESVMs as well as their promise as concerns this particular pattern recognition case