Research collaboration prediction and recommendation based on network embedding in co‐authorship networks

In large‐scale datasets, the researchers' multiple features need to be learned automatically instead of manually defined and enumerated, to improve the efficiency and effect of research collaboration prediction and recommendation. This paper applies the network embedding method to learn the context of each researcher by which the semantic similarities among researchers are calculated. Firstly, the co‐authorship network is built in a large‐scale dataset where research collaborations are denoted by co‐authorships. Then the researchers' semantic contexts in the network are learned by the network embedding method based on deep learning, and each researcher's dense, low‐dimensional vector is formed. Finally, the semantic similarities among researchers are calculated through vector similarity indices and quantitatively compared by link prediction. Experiments in the field of library and information science (LIS) verify that the method can improve the accuracy and effectiveness of research collaboration prediction and recommendation.