Personality Detection using Deep Learning Techniques

Abstract

Personality refers to the distinguishing set of qualities of an individual that impacts their attitude, habits, behaviors and pattern of thoughts. Personality traits have been shown to have governing effect on major outlook of life such as success in the political temperament, general and workplace emotional stability. Textual data accessible on Social Networking sites yields an opportunity to automatically identify personality traits of an individual. Since technology has progressed expeditiously, personality detection has become a popular research field that bestows personalization to users. Presently, researchers have employed data on social media for automatic prediction of personality. However, the extraction of the social media data is a complex process as it is noisy, available in different formats and lengths. This research proposes a machine learning model and a deep learning model to predict the personality of an individual based on Myers–Briggs Type Indicator (MBTI) personality model. The proposed machine learning models (SVM, LR, MLP and XGBoost) were trained on MBTI and MBTI500 datasets with imbalanced and balanced instances (using SMOTE). The proposed deep learning model was trained using CNN with GloVe word embeddings. SVM model achieved the highest accuracy of 96.81% for machine learning model on MBTI500 dataset with SMOTE. However, CNN exhibited the highest accuracy of 99.54% on MBTI dataset which supersedes the existing models.