More about HKUST
A Survey of Social Recommendation Systems
PhD Qualifying Examination Title: "A Survey of Social Recommendation Systems" by Mr. Hao GE Abstract: In the era of information explosion, recommendation systems (RSs) play an increasingly important role in proactively recommending interesting information to the users. The recommended information could be about products in online shopping malls, videos in video-sharing websites, or news in news websites. RSs have been studied extensively in both academia and industry. Traditional RSs try to predict a user's preferences on items utilizing only the users' historical interactions with the items, which are very commonly represented as a user-item rating matrix. Based on the ways predictions are made, approaches in traditional RSs can be generally classified into two categories, namely, content-based and collaborative filtering-based (CF-based) RSs. In this survey, we will review representative works in these two categories. Alongside RSs, online social media has gained great popularity in the past decade as well. People establish social connection with each other to build extremely large social networks in different online platforms, e.g., Facebook, Twitter, LinkedIn, and so on. To incorporate social information to enhance recommendation performance, social RSs have gained great development. Mostly social RSs are based on social influence and homophily theory, which states that socially connected users share similar preferences and influence each other??s taste in activities such as buying products or watching movies. Based on how social information is used, in this survey we classify existing social RSs research into memory-based and model-based. Due to the fact that Matrix Factorization (MF) is the most widely used method in recommendation research. Thus, within model-based social RSs, we further review three different ways to incorporate social information in MF model. From this survey, we can see that traditional RSs can be significantly enhanced by considering social connections among users, and that users?? social information is of great value in boosting performance of other machine learning tasks as well. Date: Monday, 17 June 2019 Time: 11:30am - 1:30pm Venue: Room 3494 Lifts 25/26 Committee Members: Prof. Dik-Lun Lee (Supervisor) Dr. Wilfred Ng (Chairperson) Dr. Yangqiu Song Dr. Ke Yi **** ALL are Welcome ****