AraBig5: The Big Five Personality Traits Prediction Using Machine Learning Algorithm on Arabic Tweets

Personality trait prediction has attracted substantial academic interest in recent years as a consequence of its ability to characterize people's distinctive personality features that distinguish them from others. These features help to anticipate how a person will interact with others in various situations and aid psychologists and therapists in building a profile of their clients. Personality traits are often measured through a traditional pencil-and-paper questionnaire. This questionnaire is considered expensive in terms of time and effort. Therefore, researchers have started using personality trait prediction that automatically analyzes users' social media textual data by applying machine learning techniques. However, the majority of these studies have been predicted from English texts; only one study has predicted personality traits from Arabic (Egyptian dialect) texts. Furthermore, there is a significant gap between studies that predict personality from English and Arabic texts in terms of machine learning techniques and the dataset used. This study aims to reduce the gap by building a new Arabic dataset with Saudi dialect (AraBig5) and implementing different machine learning models on AraBig5 to predict Big Five personality traits. The results indicate that logistic regression and support vector machine yielded the best average F1 scores of 0.86 and 0.87, respectively.