Multidimensional data clustering based on fast kernel density estimation

Density based clustering is an important clustering method. This paper presents a novel multidimensional data clustering algorithm based on SBR-KDE, which is a fast kernel density estimation algorithm we developed using sparse Bayesian regression, independent component analysis and data gaussianization. A pruning process of the Delaunay triangulation is also exploited in the clustering algorithm. Experimental studies using practical data and artificial data show the effectiveness of our clustering algorithm.