Personality Traits Prediction Based on Users’ Digital Footprints in Social Networks via Attention RNN

Abstract

With the increasing popularity of social networks, massive digital footprints of individuals in online service platforms are generated. As a result, an emerging technology namely personality trait analysis has drawn much attention. The prediction and analysis of personality trait is an efficient way to voting prediction, review analysis, decision analysis and marketing. The existing studies generally employ classification models while ignore the temporal property of digital footprints, which may lead to unsatisfactory results. To make an improvement, this paper proposes an effective method to predict the personality traits by taking the temporal factors into account through the use of Attention Recurrent Neural Network (AttRNN). The experimental results based on the dataset of 19000 Facebook volunteers suggest the proposed method is effective for predicting personality traits.