Exploring the significance of human mobility patterns in social link prediction
KAUST DepartmentComputer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division
Computer Science Program
Machine Intelligence & kNowledge Engineering Lab
Permanent link to this recordhttp://hdl.handle.net/10754/564840
MetadataShow full item record
AbstractLink prediction is a fundamental task in social networks. Recently, emphasis has been placed on forecasting new social ties using user mobility patterns, e.g., investigating physical and semantic co-locations for new proximity measure. This paper explores the effect of in-depth mobility patterns. Specifically, we study individuals' movement behavior, and quantify mobility on the basis of trip frequency, travel purpose and transportation mode. Our hybrid link prediction model is composed of two modules. The first module extracts mobility patterns, including travel purpose and mode, from raw trajectory data. The second module employs the extracted patterns for link prediction. We evaluate our method on two real data sets, GeoLife  and Reality Mining . Experimental results show that our hybrid model significantly improves the accuracy of social link prediction, when comparing to primary topology-based solutions. Copyright 2014 ACM.
Conference/Event name29th Annual ACM Symposium on Applied Computing, SAC 2014