harmony search algorithm for clustering with feature selection

Abstract

This paper presents a new clustering algorithm, called IHSK, with feature selection in a linear order of complexity. The algorithm is based on the combination of the harmony search and K-means algorithms. Feature selection uses both the concept of variability and a heuristic method that penalizes the presence of dimensions with a low probability of contributing to the current solution. The algorithm was tested with sets of synthetic and real data, obtaining promising results.