Facilitating social activity organization by mining human interaction history

Nowadays, people in the physical world are getting involved into various kinds of social activities, such as meetings, discussions, and journeys, etc. Followed by this is the increase of user's contacts list and the high cost on the social activity organization and activity group formation. In order to improve the efficiency of social activity organization and promote the interaction between people, we propose a mobile social activity support system based on social topology mining of the interaction history and mobile sensing. Two methods are used to recommend friends (known friends and new friends) for a certain activity. For the known-friend recommendation approach, people can get group recommendation by analyzing the interaction history and real-time contexts. For the new-friend recommendation approach, we detect the community structure of the social topology and calculate user similarity to recommend new friends. The algorithms for group formation and group recommendation are presented. We have also implemented a prototype system based on it. Experimental results verify the effect of the proposed approaches.