Improve the Performance of Link Prediction Methods in Citation Network by Using H-Index

摘要

Citation networks are widely used in graph mining. Link prediction which is the task of mining the missing links in networks or predicting the next node pair to be connected by a link is a useful tool for mining information in citation networks. Most existing studies of link prediction commonly use degree rather than more advanced methods to measure the importance of a node. However, in reality, the way to measure the importance of a paper is not that simple. Some paper have high degree in citation network but is not very influential. This issue restricts the performance of link prediction methods applying in citation network. In this paper, we use the H-index which is an advanced method to measure the importance of a paper and to enhance three classical link prediction methods. Experiments on real citation networks demonstrate that using the H-index instead of degree can performs at a higher prediction accuracy in citation network.