ComPath: User Interest Mining in Heterogeneous Signed Social Networks for Internet of People

Abstract

The Internet of People (IoP) is a human-centric computing paradigm, where the people are not considered merely as end users, but become the center of the computing architecture. The computing model of IoP requires that the system understand the social characters of the users, such as the users' emotions, personality types, and interests. User interest detection is an important task in IoP. In this article, we propose a user interest detection framework for user interest detection in the context of a signed social network for IoP. First, we propose a new proximity function that measures the similarity between users based on their interests/disinterests with respect to the relative popularity of these interests/disinterests among other users. Second, we propose a greedy community detection algorithm that detects communities of users with common interests with possible overlapping communities using the adaptive clique relaxation technique. Finally, we introduce a novel link prediction algorithm named ComPath that leverages the community affiliation information to predict the unknown links in heterogeneous signed social networks. Experimental results show that ComPath outperforms other computational-based baselines as well as deep-learning-based baselines especially in the cold start phase with only a few training data.