Skip to Main Content
Social networks have been evolving over the past few years, leading to a rapid increase in the number and complexity of relationships among their entities. In this paper, we focus on a large scale dataset known as the Digital Bibliography and Library Project (DBLP), which contains information on all publications that have been published in computer and information science related journals and conference proceedings. We model the DBLP dataset as a social network of research collaborations. DBLP is a structured and dynamic dataset stored in the XML file format; it contains over 850,000 authors and 2 million publications and the resulting collaboration social network is a scale-free network. We define DBLP collaboration social network as a graph that consists of researchers as nodes and links representing the collaboration among the researchers. In this work, we implement a data analysis algorithm called Multidimensional Scaling (MDS) to represent the degree of collaboration among the DBLP authors as Euclidean distances in order to analyze, mine and understand the relational information in this large scale network in a visual way. MDS requires a highly computational complexity for large scale graphs such as the DBLP graph. Therefore, we propose different solutions to overcome this problem, and improve the MDS performance. In addition, as the quality of the MDS result is measured by a metric known as the stress value, we use the steepest descent method to minimize the stress in an iterative process called stress optimization in order to generate the best geometric layout of the graph. We also propose a solution to further enhance the graph visualization by partitioning the graph into sub-graphs and using repelling forces among nodes within the same sub-graph.