Improving the Efficiency of a Clustering Genetic Algorithm

Finding optimal clusterings is a difficult task. Most clustering methods require the number of clusters to be specified in advance, and hierarchical methods typically produce a set of clusterings. In both cases, the user has to select the number of clusters. This paper proposes improvements for a clustering genetic algorithm that is capable of finding an optimal number of clusters and their partitions automatically, based upon numeric criteria. The proposed improvements were designed to enhance the efficiency of a clustering genetic algorithm. The modified algorithms are evaluated in several simulations.