Denny, Denny; Williams, Graham; Christen, Peter
Real-life datasets often contain small clusters of unusual sub-populations. These clusters, or 'hot spots', are usually sparse and of special interest to an analyst. We present a methodology for identifying hot spots and ranking attributes that distinguish them interactively, using visual drill-down Self-Organizing Maps. The methodology is particularly useful for understanding hot spots in high dimensional datasets. Our approach is demonstrated using a large real life taxation dataset.
Items in Open Research are protected by copyright, with all rights reserved, unless otherwise indicated.