Inferring Social Groups Using Call Logs

Recent increase in population of mobile phone users makes it a valuable source of information for social network analysis. For a given call log, how much can we tell about the person's social group? Unnoticeably, phone user's calling personality and habit has been concealed in the call logs from which we believe that it can be extracted to infer its user's social group information. In this paper, we present an end-to-end system for inferring social networks based on "only" call logs using kernel-based naive Bayesian learning. We also introduce normalized mutual information for feature selection process. Our model is evaluated with real-life call logs where it performs at high accuracy rate of 81.82%.