Impact of Person Preferences on Directed Link Prediction in Online Social Networks

Under the background of the Internet era, computational social science, emerging as the new research paradigm, can be used to test and develop the existing theoretical hypothesis of quantitative psychology. In line with this trend, our paper attempted to apply advanced computer science techniques to the analysis of personality theory for social networks. Based on the construction of social network personality prediction framework, this paper proposed a new hybrid Network Embedding model, which combined network structure information and model-based personality trait scores on the Facebook platform. Through link prediction tasks, experiments demonstrated that the hybrid model had a better prediction quality than the traditional single network embedding model. Moreover, the personality traits of “Extroversion” and “Agreeableness” were proved to be the most critical features in the link prediction problem, which further validated the key role of personality traits in social networks.