Tracing and Predicting Collaboration for Junior Scholars

Chun-Hua Tsai, Yu-Ru Lin

Abstract: Academic publication is a key indicator for measuring scholars' scientific productivity and has a crucial impact on their future career. Previous work has identified the positive association between the number of collaborators and academic productivity, which motivates the problem of tracing and predicting potential collaborators for junior scholars. Nevertheless, the insufficient publication record makes current approaches less effective for junior scholars. In this paper, we present an exploratory study of predicting junior scholars' future co-authorship in three different network density. By combining features based on affiliation, geographic and content information, the proposed model significantly outperforms the baseline methods by 12% in terms of sensitivity. Furthermore, the experiment result shows the association between network density and feature selection strategy. Our study sheds light on the re-evaluation of existing approaches to connect scholars in the emerging worldwide Web of Scholars.