Constraint-Based Clustering Algorithm for Multi-density Data and Arbitrary Shapes

Abstract

The purpose of data clustering is to identify useful patterns in the underlying dataset. However, finding clusters in data is a challenging problem especially when the clusters are being of widely varied shapes, sizes, and densities. Density-based clustering methods are the most important due to their high ability to detect arbitrary shaped clusters. Moreover these methods often show good noise-handling capabilities. Existing methods are based on DBSCAN which depends on two specified parameters (Eps and Minpts) that define a single density. Moreover, most of these methods are unsupervised, which cannot improve the clustering quality by utilizing a small number of prior knowledge. In this paper we show how background knowledge can be used to bias a density-based clustering algorithm for multi-density data. First we divide the dataset into different density levels and detect suitable density parameters for each density level. Then we describe how pairwise constraints can be used to help the algorithm expanding the clustering process based on the computed density parameters. Experimental results on both synthetic and real datasets confirm that the proposed algorithm gives better results than other semi-supervised and unsupervised clustering algorithms.