Personality Prediction using Multiple Textual Datasets and Deep Learning Models

Personality traits are a fundamental aspect of human behavior and have wide-ranging implications in psychology, marketing, and human resources. This research focuses on the classification of personality traits in human behavior by using the five-factor model. Revolutionized by advancements in computational models, this research uses Natural Language Processing (NLP) to predict human personality traits based on a variety of textual data ranging from essays to social media posts. The research proposes to determine the pattern and context using deep word embedding analysis models such as BERT and its different variants. Different preprocessing and kinds of embeddings have been performed along with thorough experimentation of the hyperparameters. Remarkably, the proposed study achieved 84.4% overall accuracy with some specific traits like cEXT being predicted at 0.97 precision and 0.88 F1-Score using a BERT model. Our multi-class classification results outperform previous results on the same datasets, highlighting the effectiveness of our proposed methodology.