Automatic extraction of common research areas in world scientograms using the multiobjective Subdue algorithm

Scientograms are graph representations of scientific information. Exploring vast amount of scientograms for scientific data analysis has been of great interest in Information Science. This work emphasizes the application of multiobjective subgraph mining for the scientogram analysis task regarding the extraction of common research areas in the world. For this task, we apply a recently proposed multiobjective Subdue (MOSubdue) algorithm for frequent subgraph mining in graph-based data. The algorithm incorporates several ideas from evolutionary multiobjective optimization. The underlying scientogram structure is a social network, i.e., a graph, MOSubdue can uncover common (or frequent) scientific structures to different scientograms. MOSubdue performs scientogram mining by jointly maximizing two objectives, the support (or frequency) and complexity of the mined scientific structures. Experimental results on five realworld datasets from Elsevier-Scopus scientific database clearly demonstrated the potential of multiobjective subgraph mining in scientogram analysis.