Histographs: Interactive Clustering of Stacked Graphs

Visualization systems must intuitively display and allow interaction with large multivariate data on low-dimensional displays. One problem often encountered in the process is occlusion: the ambiguity that occurs when records from different data sets are mapped to the same display location. For example, because of occlusion summarizing 1000 graphs by simply stacking them one over another is pointless. We solve this problem by adapting the solution to a similar problem in the Information Murals system [2]: mapping the number of data elements at a location to display luminance. Inspired by histograms, which map data frequency to space, we call our solution histographs. By treating a histograph as a digital image, we can blur and highlight edges to emphasize data features. We also support interactive clustering of the data with data zooming and shape-based selection. We are currently investigating alternative occlusion blending schemes.