On Neighborhood Effects in Location-Based Social Networks

In this paper, we analyze factors that determine the check-in decisions of users on venues using a location-based social network dataset. Based on a Foursquare dataset constructed from Singapore-based users, we devise a stringent criteria to identify the actual home locations of a subset of users. Using these users' check-ins, we aim to ascertain the neighborhood effect on the venues visited, compared with the activity level of users. We further formulate the check-in count prediction and check-in prediction tasks. A comprehensive set of features have been defined and they encompass information from users, venues, their neighbors, and friendship networks. We next propose regression and classification models to address the two prediction tasks respectively. Our experiments have shown that the two models especially the classification models outperform the baseline methods when all features are used. We also analyze feature importance and found that despite their similarity, the two prediction tasks actually require different weights on the features as learned by the regression and classification models. Finally, it was found that user's home location for deriving user-venue distance feature is a better feature than user's center of the mass.