Predicting human contacts in mobile social networks using supervised learning

Having access to human contact traces has allowed researchers to study and understand how people contact each other in different social settings. However, most of the existing human contact traces are limited in the number of deployed Bluetooth sensors. In most experiments, there are two types of participants, the ordinary ones who carry cellphones and a specially selected group who additionally carry sensors. Although the contacts between any pair of participants are known when at least one of them carry a sensor, the contacts between any pair of participants are "hidden" when both of them carry their cellphones. In this paper, we employ two well-known supervised classifiers for predicting hidden contacts among participants who carry their cellphones. The performance results of our supervised classifiers show the applicability of using machine learning algorithms for contact prediction task. The results also show that a small subset of features such as number of common neighbors and total overlap time play essential roles in forming human contacts. Finally, we show that contacts of nodes with high centralities are more predictable than nodes with low centralities.