Group Link Prediction

Due to its universal applications in the domain of social network analysis, e-commerce, and recommendation systems, the task of link prediction has received enormous attention from the data mining and machine learning communities over the last decade. In its original setting, the task only predicts whether a pair of entities who are not connected at present time will form a connection in future. However, in real-life an entity sometimes join a group (or a community), thus making a connection with the group (or the community), instead of connecting with an individual. Existing solutions to link prediction are inadequate for solving this prediction task. To overcome this challenge, in this work we propose a novel problem named group link prediction which focuses on evaluating the likelihood for a candidate to become a member of a group at a given time. The problem has potential applications such as friendship or group suggestions on Facebook or other social networks, as well as co-authorship suggestion, or group email recommendations. To solve the problem, we propose a Long Short-term Memory based model that inputs the embedding vectors of the group and outputs the conditional probability distributions for the candidates. We also introduce a composite long short-term memory model that integrates keyword information. Experimental results on real-world data sets validate the superiority of our proposed model in comparison to various baseline methods.