User-guided Graph Exploration: A Framework for Algorithmic Complexity Reduction in Large Data Sets

Tim Grube, Florian Volk, Max Mühlhäuser, Suhas Bhairav, Vinay Sachidan, a, Yuval Elovici

Human exploration of large data sets becomes increasingly difficult with growing amounts of data. For this purpose, such data sets are often visualized as large graphs, depicting information and interrelations as interconnected vertices. A visual representation of such large graphs (for example, social networks, collaboration analyses or biological data sets) has to find a trade-off between showing details in a magnified—or zoomedin—view and the overall graph structure. Showing these two aspects at the same time results in a visual overload that is largely inaccessible to human users. In this article, we augment previous work and present a new approach to address this overload by combining and extending graph-theoretic properties with community detection algorithms. Our non-destructive approach to reducing visual complexity while retaining core properties of the given graph is user-guided and semi-automated. The results yielded by applying our approach to large real-world network data sets reveal a massive reduction of displayed vertices and connections while keeping essential graph structures intact.