Interactive Visualization of Streaming Data with Kernel Density Estimation (ACM)
In this paper, we discuss the extension and integration of the statistical concept of Kernel Density Estimation (KDE) in a scatterplot-like visualization for dynamic data at interactive rates. We present a line kernel for representing streaming data, we discuss how the concept of KDE can be adapted to enable a continuous representation of the distribution of a dependent variable of a 2D domain. We propose to automatically adapt the kernel bandwith of KDE to the viewport settings, in an interactive visualization environment that allows zooming and panning. We also present a GPU-based realization of KDE that leads to interactive frame rates, even for comparably large datasets. Finally, we demonstrate the usefulness of our approach in the context of three application scenarios - one studying streaming ship traffic data, another one from the oil & gas domain, where process data from the operation of an oil rig is streaming in to an on-shore operational center, and a third one studying commercial air traffic in the US spanning 1987 to 2008.
Paper available at ACM.