Strength-Based Link Prediction in Scientific Bibliographic Networks

Abstract

Network analysis literature counts plenty of models of different paradigms designed for solving the link prediction problem in complex information networks. However, fewer studies that have exploited link strength-related social theories for this purpose even in a social context. In this paper, the authors introduce a new approach to solve the link prediction problem in scientific bibliographic networks. The aim is to predict future collaboration relations between scientists relying upon the “strength of strong ties” hypothesis. The proposed model estimates the strength of a relation between two scientists using a set of efficient link strength indicators. The importance of the relation is then validated according to the scientists' expected collaboration strategies. The prediction process is performed in a heterogeneous context where the types of the nodes and the links are considered. Experiments on the DBLP real-world scientific bibliographic network, show higher performance of our model in comparison with the link prediction baseline methods.