Evaluation of features for social contact prediction

Human contact prediction is considered a challenging problem in the context of social networks. We have approached this problem as link prediction problem and analyzed social profiles and their impact on predicting links in a social network. In this work, we have argued that not all social features share the same importance for link prediction. The experiments in this paper are performed using varying weights for the available features in the dataset. Our results show that we can gain significant improvement in link prediction while using the suggested weighted features of social profiles.