Predicting Link Sign in Online Social Networks based on Social Psychology Theory and Machine Learning Techniques

Online social networks provide a great platform for internet users to share their views and ideas. Social media provides a dynamic platform that includes the formation and deformation of connections. Two types of connections, i.e., Positive and Negative, can exist in a social network. Positive connections are a sign of friendship or trust, while negative connections show enmity or distrust. Various applications in several fields have networks containing both positive and negative edges. Reliable prediction of edge sign can greatly influence in recommending friendly relationships while preventing enemy relationships across the network. Prediction of edge signs has been explored previously also. However, we intend to predict the sign of edges based on extracted features of nodes constructed upon theories of social psychology that includes classical balance theory and the status theory. Moreover, we employ emotional information theory and use the combined extracted features from all the theories to analyze networks for better prediction. Our results show that the proposed methodology has obtained significant accuracy when implemented on two real-life datasets, namely Slashdot and Epinions.