Strategies for Visual Data Mining

Edward J. Wegman, George Mason University

The human visual system is a powerful tool for recognizing patterns in data displays, hence, as a tool for hypothesis generation. There are substantial issues associated with the visualization of large-scale, multidimensional data. We discuss some visualization tools including parallel coordinates, grand tour, and saturation brushing and illustrate how we use these tools for exploratory discovery. Some strategies we employ are the so-called BRUSH-TOUR and TOUR-PRUNE strategies. Recently, there has been a recognition that approaches for the analysis of streaming data need development. We also suggest some tools for streaming data.