Effective social relationship measurement based on user trajectory analysis

Social community structure is widely utilized in the study of modeling disease propagation, information dissemination and etc. Therefore, detecting the social community structure is one of the most significant tasks for majority of existing works focusing on the online social network. Traditionally, they aim at predicting the existence of social relationships based on cyber interactions (e.g. online conversation) among users. However, the strength information of social relationships is not captured which is as important as the topology information of social communities. Furthermore, physical interactions (e.g., face to face conversation), which have the potential to reflect more realistic state of social relationships than cyber ones, are not taken into account in social relationship measurement. In order to measure the strength of social relationships, in this paper, we propose a hierarchical entropy-based relationship measurement approach (HERMA). HERMA is able to measure the strength of social relationships among users based on their physical interactions which could be inferred by analyzing co-location records extracted from their trajectories. To model users’ co-location records in HERMA, a hierarchical region structure is designed. Moreover, two novel concepts called user entropy and area entropy adopted by HERMA are proposed to quantify the activeness degree of an user and the openness degree of an area respectively. Finally, to validate the effectiveness of HERMA, simulations are conducted of which the results show that HERMA outperforms the baselines by leveraging the highest average accuracy on the measurement of social relationships.