Toward local family relationship discovery in location-based social network

Abstract

The local family relationship discovery problem in location-based social network (LBSN) services is to identify whether two local residents in a city belong to the same family or not by using their check-in traces on LBSNs. This information is critical for many applications, such as social relationship analysis, targeted ads of local businesses, census study, localized news and travel recommendations. In this study, we propose an unsupervised approach to solving the local family relationship discovery problem by exploiting spatial–temporal, categorical and social constraints from the noisy LBSN data. The spatial–temporal constraint represents the correlations between people and the venues they visit, the categorical constraint represents the category of the visited venues and the social constraint represents the social connections between people. In particular, we develop a local family relationship discovery (LFRD) framework that contains two major components: (1) a localness-aware expectation maximization scheme to correctly identify the local residents in a city and (2) a family relationship discovery scheme to discover family relationships between the identified local people. We study the performance of the LFRD framework using four real-world datasets collected from Foursquare. The LFRD is shown to outperform the state-of-the-art baselines by significantly improving the accuracy of family relationship discovery.