Citation Networks as a Multi-layer Graph: Link Prediction and Importance Ranking

ABSTRACT In academia, to represent the relationships between researchers or papers, the traditional way is to analyze the citation network constructed by paper citation relationships. Our major contribution is that we incorporated multiple networks from the publication dataset, such as the paper citation network, author citation network, author collaboration network, etc. These networks can provide very useful and interesting information for analyzing the citation behavior of researchers. In this paper, we first analyze how people are citing papers by applying logistic regression model to paper citation network. With the additional information from the author networks, we can get improved results of link prediction in citation network. Then we run a modified PageRank algorithm on paper citation network and author citation network, which can provide us with a more accurate sense of the paper and researcher impact.