Serendipitous User Recommendation in Twitter by Consider Unexpected and Useful Interests

Abstract

　Many user recommendation systems in Twitter mainly focus on prediction accuracy, making recommendations based on users' profiles. However, such recommendations tend to stay within the scope of the users' existing interests. Consequently, it becomes difficult to introduce new discoveries or surprises, which makes it challenging to improve users' satisfaction. To solve this problem, we propose a method to recommend serendipitous users for Twitter by consider unexpected and useful interests. The effectiveness of the recommender system was evaluated by comparing it with existing methods. The results showed that our proposed method recommended more useful users, as well as recommending more serendipitous users to experimental participants in some cases.