New methods for calculating trend-based vertex similarity for collaboration recommendation

Abstract

Collaboration recommendation is a problem that automatically selects and provides a list of potential researchers or research groups with respect to the input which is a researcher or a research group. Recently, this problem has attracted a lot of attention of many researchers in this area. A popular approach for collaboration recommendation problem is based on social network analysis, specifically co-author network analysis. However, the current methods do not consider collaborative trend in analyzing co-author network and the collaborative trend is one of the key factors for forming new co-authorships. In this paper, we propose three new methods: (1) Maximum Path based Relation Strength (MPRS); (2) Maximum Path based Relation Strength Plus (MPVS+); and (3) Relation Strength Similarity Plus (RSS+), for modeling and calculating vertex similarity in the co-author network. In our trend-based methods (MPVS+, RSS+), information of collaborative trend is used to improve the calculation of relation strength for researchers in the co-author network. The proposed methods are applied for researcher collaboration recommendation. Experiments are conducted on two dataset: i) Digital Bibliography & Library Project (DBLP), one popular and public science database; ii) the dataset extracted from the Microsoft Academic Search[1] website. The experiment results show that our proposed methods are more effective than the existing vertex similarity methods in predicting co-author collaboration.