Personality Modelling of Indonesian Twitter Users with XGBoost Based on the Five Factor Model

Abstract

Over the past few years, there has been an increasing number of researches on automated personality prediction. One of the approaches include analysing personality based on the user’s choice of words on social media. Even though research on personality prediction have been done in several different languages, advancements in personality prediction for Bahasa Indonesia, the Indonesian language, has been stagnant. This is due to scarcity of data and the lack of psycholinguistic dictionaries for the language. This is unfortunate as Indonesians are among one of the most active social media consumers, considering more than 2% of worldwide tweets come from Indonesians. We address these issues in this study through the modelling of our own personality prediction model based on an Indonesian dataset that we have produced. We employed the model using the XGBoost machine learning algorithm, trained on 250 user data that we have collected and annotated manually. The resulting model was able to gain decent performance for the Agreeableness and Openness personality trait, achieving AUROC of 0.71 and 0.63, while the Conscientiousness, Extraversion, and Neuroticism personality traits were harder to distinguish with an AUROC of 0.5, 0.59, and 0.48 respectively.