Social Network Centrality Measures for Ranking Academic Authors

Social Network Centrality Measures

Abstract

The progression in the arena of IT particularly in IoT, social media and communication technologies, knowledge turn out to be available on finger tips in every field of life. Perceiving the world through the lens of overlapping networks that transfer information, knowledge, and power reveals insights into trends, technology, and interests. Analytics play vital role in the field academic network research and development. Analyzing social academic networks provides new perspectives on various interesting topics. The social connections have a significant impact on our actions, thoughts, and knowledge. However, standard statistical methods lack a reliable method for considering the impact of strong connections. This can only be achieved through academic network analysis and by comparing and contrasting relevant data. The social networking data analysis gives us tools to quantify the social network connections. Centrality can be used to quantify a node's importance and influence within the network as a whole. The concept of importance has various implications depending on the type of network being analyzed. Centrality indices ask the question, "What characterizes the significance of a node?" Different centrality measures can be used to demonstrate a node's importance. In this study, we analyzed and experimented various network centrality methods, their characteristics, and limitations using S2ORC dataset that consisting of 81 million heterogeneous academic objects with over 136 million nodes. The results have listed the top influential authors by using the Network Centrality measures such as an author Anand Radhakrishnan ranked 5th with a value of 30, 3^rd with a value of 16.1 and 15th with a value of 0.386 in Degree, Betweenness and Closeness Centralities, respectively. The results are tested using performance matrices such as Spearman Correlation, Kendall Correlation and Similarity. The result of all the measures were consistent with each other. This study will also helpful in future researches to measure the semantic ranking of authors.