Enhancing MBTI Personality Prediction from Text Data with Advance Word Embedding Technique.

Abstract

Understanding human personality traits is crucial for various domains, including psychology, education, and human resources. The Myers-Briggs Type Indicator (MBTI) is a widely recognized psychological assessment tool, categorizing individuals into one of sixteen distinct personality types. The existing methodologies, which primarily relied on Word2Vec embeddings and traditional machine learning models, showed promise but left room for improvement. To address this problem, this research focused on enhancing Myers-Briggs Type Indicator (MBTI) personality prediction from text data through advanced word-embedding techniques, specifically GloVe and BERT. The research investigates the effectiveness of various Machine Learning Classifiers, including Random Forest, XGBoost, LinearSVC, SGD, Logistic Regression, and CatBoost, in predicting MBTI personality types. Additionally, the impact of preprocessing techniques such as text cleaning, tokenization, TF-IDF vectorization, GloVe, and BERT embeddings on classification performance is examined. Furthermore, the research explores strategies for addressing class imbalance through upsampling techniques. Results indicate high accuracy and performance across multiple classifiers, with XGBoost achieving the highest accuracy of 97.33%. The analysis of MBTI dimensions reveals nuanced insights into the classifiers' ability to capture specific personality traits.