Improving of Initial Clusters Fitness in Genetic Guided-Clustering Ensembles

Abstract

The clustering ensemble is a new topic in machine learning. It can combine multiple partitions generated by different clustering algorithms into a single clustering solution. Genetic algorithms have been known as methods with high ability to find the solution of optimization problems like the clustering ensemble problem. So far, many contributions have been done to find consensus cluster partition by genetic algorithms; however there has been little discussion about the methods of carrying out the initialization population and generation of initial cluster partitions in the first phase of clustering ensembles. In this paper, we proposed a new algorithm that used by clustering ensembles which improve cluster partitions fitness. In addition, diversity clustering problem has been solved by used the proposed algorithm. We compared the fitness average among individuals generated by the proposed algorithm and other clustering algorithms which have been calculated by three different fitness functions. The obtained experimental results on several benchmark datasets have demonstrated the proposed algorithm improve cluster solutions fitness.