Geo-Social Ranking: functions and query processing

Abstract

Given a query location q, Geo-Social Ranking (GSR) ranks the users of a Geo-Social Network based on their distance to q, the number of their friends in the vicinity of q, and possibly the connectivity of those friends. We propose a general GSR framework and four GSR functions that assign scores in different ways: (i) LC, which is a weighted linear combination of social (i.e., friendships) and spatial (i.e., distance to q) aspects, (ii) RC, which is a ratio combination of the two aspects, (iii) HGS, which considers the number of friends in coincident circles centered at q, and (iv) GST, which takes into account triangles of friends in the vicinity of q. We investigate the behavior of the functions, qualitatively assess their results, and study the effects of their parameters. Moreover, for each ranking function, we design a query processing technique that utilizes its specific characteristics to efficiently retrieve the top-k users. Finally, we experimentally evaluate the performance of the top-k algorithms with real and synthetic datasets.