Can multiple social ties help improve human location prediction?

Abstract

Predicting the future locations of human is of vital importance in practical applications, such as the urban planning and epidemic controlling. Here we explore how to improve the location prediction accuracy by utilizing multiple human social ties. Besides the trajectories of individuals themselves, the dense social tie, i.e friends and in-roles (F&IR), and the familiar stranger (FS) social tie are also involved to design basic location predictors, according to the social-tie properties. Interestingly, we find that the human location predictor with the FS has an even higher prediction accuracy than the predictor with the F&IR social tie. Moreover, we attempt to select a proper predictor for humans with heterogeneity. We define a social-tie motif to validate the social-tie pattern of an individual and propose a multiple-social-ties (MST) location predication method. The MST predictor opens up avenues for the location prediction considering the social-tie patterns of individuals.