Skip to Main Content
One of the first questions we pose when we have a new data set to explore is: are there any obvious groupings in the data? These clusters reveal the structure of the data set and make the data easier to understand. Cluster analysis is a classic problem and there are many tools for it. The problem is that these methods often give conflicting and too fine-grained results for practical use. As well, they are accompanied with the danger of over-interpreting the data. Another drawback in these techniques is that they do not usually provide an opportunity for interactive exploration. We propose a simple user interface and method to interactively explore clusters in data sets. The method is based on a rapid dynamic query and it can detect both hierarchical and partitional clusters.
Date of Conference: 14-16 July 2004