A visual recommendation system for co-authorship social networks (ChinaVis 2018)

User recommendation plays a crucial role in social network applications such as co-authorship networks. Existing techniques mostly strive to pursue the similarity between nodes or the accuracy of link prediction, leading personal networks to be monotonic. However, users often have various expectations regarding the growths of their social networks, which is clearly hard to capture via automatic algorithms. In addition, adopting a recommendation likely introduces subtle changes to a social network, which may further influence the next stage of recommendation. These highly personalized and dynamic aspects of the growth of a personal social network are rarely touched in existing work. In this project, we introduce an expectation-driven visual recommendation system to address the customized demands in co-authorship social networks. The system characterizes a person’s social network with the tags of the friends. It visually presents the changes made by individual recommendations, including the direct changes to the network and the potential changes that will be introduced by possible subsequent recommendations. A visual simulation interface allows users to add friends from the recommended list. The recommendation result will be updated instantly for inspection. Thus, users can comprehensively compare different growth strategies to find the most beneficial one. We demonstrate the system with DBLP academic co-authorship network to confirm its effectiveness and efficiency.