A Particle Swarm Optimization derivative applied to cluster analysis

Modern machine learning and data analysis hinge on sophisticated search techniques. In general, exploration in high-dimensional and multi-modal spaces is needed. Some algorithms that imitate certain natural principles, the so-called evolutionary algorithms, have been used in different aspects of Environmental Science and have found numerous applications in Environmental related problems. In this paper we apply a derivative of PSO (Particle Swarm Optimization), recently introduced by the authors to partitional clustering of a real-world data set obtained from a Water Supply Company. The PSO derivative we consider here improves several typical features of this optimization technique. For one thing, PSO is adapted to consider mixed discrete-continuous optimization since the problem we tackle here involves the use of both continuous and discrete variables. For another, one of the main drawbacks associated with PSO comes from the fact that it is difficult to keep good levels of population diversity and to balance local and global searches. This formulation is able to find optimum or near-optimum solutions much more efficiently and with considerably less computational effort because of the richer population diversity it introduces. Finally, the cumbersome aspect, common to all metaheuristics, of choosing the right parameter values is tackled through self-adaptive dynamic parameter control