RoBERTa-Large LLM for Examining Mental Health of Students through Sentiment Analysis and Data Analytics

Hikmat Ullah Khan^1*Anam Naz¹Fawaz Khaled Alarfaj^2*Naif Almusallam²

• ¹University of Sargodha, Sargodha, Pakistan
• ²King Faisal University, Al-Ahsa, Eastern Province, Saudi Arabia

Mental health of students plays an important role in the overall well-being as well as their academic performance. The growing pressure from academic concerns, co-curriculum activities such as sports, and personal challenges demand modern methods for monitoring mental health. Traditional approaches depend on self-reported surveys and psychological evaluations, which can be particular and time-taking. With the progression of Artificial Intelligence (AI) particularly in Natural Language Processing (NLP), sentiment analysis has emerged as an effective technique for detecting mental health patterns through textual data. However, analyzing students' mental health remains a challenging task due to the intensity of emotional expressions, linguistic variations, and context-dependent sentiments. In this study, our main aim is to examine the mental health of students by carrying out sentiment analysis using advanced deep learning models. To accomplish this task, the state-of-the-art Large Language Model (LLM) approaches such as RoBERTa (Robustly optimized BERT approach) Large and ELECTRA have been used for empirical analysis. RoBERTa-Large, based on Google based BERT, is a larger architecture which captures more complex patterns and performs better on various natural language processing tasks. A few deep learning algorithms have been applied and RoBERTa-Large achieves the highest accuracy of 97% while ELECTRA shows 91% results on multi-classification task showing seven diverse mental health status class labels which help to predict the mental health of all those individuals who are involved in physical activities.