Clustering Methods for Interval-Valued Data

• Activity Number: 613
• Type: Contributed
• Date/Time: Thursday, August 7, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Computing
• Author(s): Yi Chen*+ and Lynne Billard
• Companies: University of Georgia and University of Georgia
• Keywords: symbolic data ; interval-valued data ; divisive monothetic clustering ; Hausdorff distance

Abstract

With the development of computing and internet technology, datasets with stupendously large number of observations are more and more common. One technique to handle the big data is to aggregate classical data to symbolic data, like lists, intervals, lists with probabilities and intervals with probabilities (histograms). Building clustering methods for symbolic data has been an active area over the past decade. In this paper, we first review the divisive monothetic clustering method for interval data proposed by Chavent (1998, 2000). Then, the algorithm is implemented in the software package R, and comparisons among different Hausdorff distances are made. Finally, the method is applied to the practical data. Advantages and disadvantages of using different distances for clustering are also discussed.