MBTI Personality Prediction Based on BERT Classification

Young people today tend to express their feelings and socialize on the internet instead of in real life, which makes social media practical in defining one's personality since their expressions usually exhibit their personalities. Predicting people's personalities based on their posts is a relatively challenging task requiring large quantities of processing data and modeling. This paper uses two word-embedding methods, BERT classification and TF-IDF Vectorizer, and three models, including Logistic Regression, K-Nearest Neighbors, and Random Forest Classifier, to find this task's state of the art method. In this case, with BERT classification, the state-of-the-art method for most of the Natural Language Processing(NLP) tasks, Logistic Regression is the best-performing model with an average accuracy of 87 percent.