Pairwise influence based on spatio-temporal data

Abstract

With the popularity of smartphones and social network services (SNS), online relationships take up much of our lives, but traditional offline relationships are still important. Recently, large-scale spatial-temporal data is generated through the movement of people in the real world through the spread of GPS-embedded devices and Location Based Service (LBS). We try to find and quantify offline relationships from spatial-temporal data. We propose a model consisting of temporal pairwise and spatial influences. Temporal pairwise influence infers and quantifies offline relationships through inductivity and followship. Spatial influence solves the coincidence problem by inferring the characteristics of the place through Location Entropy (LE) and Personal Meaning of a Location (PML). Location Entropy infers to the characteristics of a place in society, and Personal Meaning of a location infers the meaning of a place to an individual. We verified the proposed model in two ways. In the comparative experiment with online relationships to show the reliability of the model, our model showed more than 90% similarity. We have experimented with local-based networks to show that the need to consider local features and that our model reflects local characteristics. Our model not only analyzes real-world human relationships through spatial-temporal data, but also can be an alternative to human relationship analysis that relies on online social networks. It is also expected that the combination of our model and existing online human relations studies will allow us to deepen our understanding of modern people's relationships.