A graph based data mining method for collaborative learning space in learning commons

Abstract

A graph based data mining method, which discovers automatically usage patterns from user-to-user and user-to-object interactions in a collaborative learning space, is proposed. The proposal describes mathematically observed users, objects, and their interactions at a given time as a set of graphs (a usage pattern) whose node is a user or an object and edge is assigned depending on a physical distance between two nodes. It is validated that the proposal can provide useful data for interview planning and evidences for interview results. On the validation, detection of frequent local usage patterns, detection of rare spatial layouts among usage patterns, and grouping hours containing similar local usage patterns are demonstrated with the 324 pictures taken at the collaborative learning space in Chiba University Library.