An Enhanced Approach for Quantitative Prediction of Personality in Facebook Posts

Social media is a collection of computer-mediated technologies that encourages the creation and sharing of data, thoughts and vocation interests by means of online communities. There are various kinds of web-based social networking i.e. micro-blogs, wikis and social networking sites. Different social media like Facebook, LinkedIn, Google+ and Twitter are the popular sources for connecting people all over the globe. Facebook is one of the commonly used platform where individual’s used to stay in touch, business personnel used for marketing and others used to share expedient information. Due to this lucrative nature, one’s personality can be predicted on the basis of posts created, commented on others post and likes against any posts. We have developed in-house tool using python language that defines personality in terms of psychological model of Big-5 personality traits including extraversion, neuroticism, agreeableness, openness and conscientiousness. The dictionary based approach has been used in this tool in which we have combined three dictionaries (WordNet, SenticNet and Opinion Lexicon). Our proposed technique has shown promising results as we have analyzed 213 unique Facebook profiles and their results outperforms the others. Furthermore a comparative analysis of machine learning classifiers i.e. support vector machine, naïve bays and decision tree has performed. Our approach succeeds to predict personality traits. We are intended to predict personality from roman English posts in future.