Personality assessment from social media data: An ensemble model

Abstract

The global prevalence of social media encourages people to upload and share a vast and recurrent amount of information about themselves through various mediums of communication such as text, pictures, audio, and video. These means of communication are embedded with people's interests, emotions, values and personality that can be collected from this data. This significant amount of information has become the interest of different professionals and industries such as data scientists, psychologist, businesses, etc., primarily to predict the behavior, traits and potential interests of people on online platforms to offer various products or services to users geared towards their benefit. Our research findings reported in this thesis indicate that the personality of a user can be assessed through analyzing their photos on social media, and is supported by a review of relevant literature published since 1999 and a comparative study and performance analysis using evaluation metrics over twelve state-of-the-art research studies published from 2016. As a result, this thesis introduces a new ensemble model that achieves the improved accuracy for each personality attire and shows that using an optimized feature space improves prediction performance.