MBTI PERSONALITY PREDICTION USING TRANSFORMER BASED DEEP LEARNING MODELS

Personality prediction plays an important role in areas such as recruitment, education, healthcare, and behavioral analysis. Traditional personality assessment methods mainly depend on questionnaires, which are often time-consuming, biased, and difficult to scale for large populations. With the rapid growth of social media platforms, people continuously express their thoughts, emotions, and opinions through text, providing valuable data for automatic personality analysis. This project presents an automated MBTI personality prediction system using social media text and transformer-based deep learning techniques. A dataset containing social media posts of 8,675 users. The textual data was first preprocessed by removing noise, links, punctuation, and unnecessary symbols to improve data quality. Multiple machine learning and deep learning models were explored, including SVM, Random Forest, LSTM, BiLSTM, DistilBERT, and RoBERTa. A hybrid approach was adopted in which contextual embeddings generated by transformer models were integrated with LSTM architectures for enhanced personality prediction. The proposed system was evaluated on binary classification, four-dimensional MBTI classification, and full 16-personality-type prediction. Experimental results showed that the hybrid model significantly outperformed traditional methods, achieving high accuracies. The findings confirm that social media text contains rich personality-related information.

Poster