“…In some cases, the interactivity is justified by computational requirements-any visualization that takes too much time and computational power to generate is more or less useless in an interactive setting, thus methods to speed up the operation (through optimization, approximation, or pre-fetching results) fit here. For example, Ahmed and Weaver [1] describes a heuristic pre-fetching method that leads to improved response time of the interactive k-means algorithm in "dynamic query visualizations of multidimensional data" ([1], p. 1). Rasmussen and Karypis [82] provides a clustering platform that allows running multiple clustering algorithms with a work-flow, i.e., import and prepare data, select clustering options, generate reports, and display visualization.…”