Predicting Personality Traits of Users in Social Networks

Abstract

With the popularity of social media, an emerging interest is on predicting personality by mining users’ digital footprints. Some researches have proven that social behaviors of users in social network are strongly influenced by their personality. However, current methods are mainly focused on selecting better features from user behaviors and then utilizing a classical classification model to separately predict different personality traits. Based on our observation and statistic analysis from the data, this paper proposes a unified semi-supervised method, Personality-dependent Variable Factor Graph model (PVFG), by considering not only the interrelations between a person’s personality traits, but also the interrelations with their friends’ personality traits. The experiment is carried on two real-world datasets and the results indicate that the proposed method outperforms several alternative methods.