An Effective Personality Prediction Using Bidirectional Encoder Representations from Transformers with AdamW Optimizer

In recent days, each and every individual personality reflects the individual behavior of person. Therefore, it is essential to assess the personality that enhances the behavior of individuals across the world. Previous researchers have used various machine learning and deep learning analyzes individuals' traits across psychology, human resources, marketing, and mental health. However, predicting personality is challenging due to the subjective nature of traits, ethical considerations around data privacy, and the need for models that work accurately across diverse backgrounds. This study suggests a novel approach to develop a deep learning model that resolves these challenges by combining various techniques such as Bidirectional Encoder Representations from Transformers (BERT), AdamW Optimizer (BERT-AdamW). Initially, the data is collected from the Myers-Briggs Type Indicator (MBTI) dataset. Further, the data is preprocessed using lowercasing, lemmatization, and stemming and processed through a pre-trained language model namely BERT for feature extraction of optimal features. In addition, the AdamW optimizer is utilized to enhance the performance of BERT and finally LR model is incorporated to predict the personality behavior of an individual. The BERT-AdamW model resulted with 97.89% accuracy when compared with Long Short-Term Memory (LSTM).