Visualizing Sequential Educational Datamining Patterns

The use of educational datamining techniques to elicit patterns in student behavior and learning outcomes can be a useful for the analysis of the effectiveness of teaching strategies and the coordination of study programs. However, results from those techniques are, often, large sets of symbolic patterns, numbering in the thousands, usually presented in text format. This makes them hard to understand which, coupled with the lack of an overall view, hinders a more comprehensive data analysis. The authors propose that information visualization techniques can be used to display relevant information in those patterns in effective ways, allowing decision makers to better insights about the reality at hand. They present a solution built upon two linked views, one based on node-link representations and another a multi-matrix representation. The complementarity of both visualization techniques allows the most important patterns to be immediately apparent, while at the same time permitting their interactive exploration in meaningful ways. The authors performed user tests proving their effectiveness.