Cluster Rendering of Skewed Datasets via Visualization.pdf

Cluster Rendering of Skewed Datasets via Visualization Keke Chen Ling Liu College of Computing, Georgia Institute of Technology 801 Atlantic Drive Atlanta, GA 30332 Tel: 1-404-894-7008, 1-404-385-1139 {kekechen, lingliu}@ ABSTRACT Information Visualization is commonly recognized as a useful method for understanding sophistication in large datasets. In this paper, we introduce a flexible clustering approach with visualization techniques, aiming at the datasets that have skewed cluster distribution. This paper has three contributions. First, we propose a framework Vista that incorporates information visualization methods into the clustering process in order to enhance the understanding of the intermediate clustering results and allow user to revise the clustering results easily. Second, we develop a visualization model that maps multidimensional dataset to 2D visualizations while preserving or partially preserving clusters. Third, based on the visualization model, a set of operating rules are proposed to guide the user rendering clusters efficiently. Experiments show that the Vista system can yield lower error rates for real datasets than typical automated algorithms. Keywords Information visualization, clustering, interactive techniques 1. INTRODUCTION Clustering is a common technique used in understanding and manipulating datasets, such as bioinformatic datasets and high- energy physical datasets. Many clustering algorithms have been proposed. Most of them automate the entire clustering process. What user needs to do is to set parameters at the beginning, such as the number of representative points and shrink factor in CURE [3]. Even though most of the algorithms allow people to use different input parameters, once the input arguments are set, the clustering result is produced with no interruption, leaving less flexibility to incorporate any application-specific revision into the clustering rules. Many of them are also