Cluster Analysis on High-Dimensional Data: A Comparison of Density-based Clustering Algorithms

The effectiveness and efficiency of the existing cluster analysis methods are limited, especially when the referred data has high dimensions or when the clusters within the data are not well-separated and having different densities, sizes and shapes. Density-based clustering algorithms have been proven able to discovered clusters with those characteristics. Previous researchers which explored density-based clustering algorithms focused on the analyzing the parameters essential for creating meaningful spatial clusters. The aim of this paper is to provide a comparative study of three well know density-based clustering algorithms including DBSCAN, DENCLUE and LTKC. The merits of them were evaluated of their ability to cluster several high-dimensional artificial data. We concluded that each density-based data clustering algorithm has their individual merits for highdimensional data. However, further research is needed in the application of the techniques to analyze other high-dimensional data, to permit a comprehensive evaluation of their respective strengths and limitations as powerful cluster analysis methods.