Development of new evolutionary schemes for clustering-like problems

Abstract

In this Ph.D. thesis we propose a general theoretical framework for dealing with clustering-like problems. The application of the theoretical framework using evolutionary algorithms to specific problems is presented. We focus on the evolutionary solution of the problems, as general methodology which can be also adapted to the new definitions and concepts described in the theoretical framework proposed. We present the solution of three clustering-like applications, with real applicability in Engineering and Science fields: the optimal partition of an Industrial Ethernet network, the color reduction problem in images (with direct application in image compression), and finally the wind speed reconstruction in wind farms, from synoptic pressure fields (key in the previous studies for installing wind farms). All the problems are fully described in terms of the theoretical framework proposed, and successfully solved by using evolutionary computation algorithms, obtaining improvements over alternative solution methods.