Personality Estimation using Demographic Data in a Personality-based Recommender System: A Proposal

Abstract

Collaborative filtering in a recommender system has a weakness called cold start problem. One way to resolve this problem is by using personality traits that can be automatically predicted from the status that the users write in social media like Facebook and Twitter. The problem with this method is that a user of such system must have at least one account in at least one social media and must write at least one status with certain length. We propose to use the combination of personality traits and demographic data to overcome this problem. Previous studies reveal that personality traits are influenced by age and gender. By using these findings, we will build models to predict personality traits from such demographic data. The modeling will be conducted by means of classification and association rule methods. Novel domains will be used in the proposed system, namely sports and hobbies.