Abstract: In order to mining the social relationship between users based on spatio-temporal data, a novel entropy-personal-time-duration-diameter (EPTDD) model is proposed for measuring relationship strength in this paper. The model considers the effect on relationship measurement of meeting events from several different sides including location, time and user on the basis of merged meeting events. Firstly, meeting events are merged according to their occurring times to obtain merged meeting events that are more correlated with real life. Each merged meeting event is then weighted from location entropy factor, location personal factor, time factor, duration factor and diameter factor. Finally, the five factors are synthesized to obtain unsupervised and supervised methods for measuring social relationship. Experimental results on three different real datasets demonstrate that our methods perform significantly more favorable than existing methods on the effectiveness.