MINING SOCIAL MEDIA FOR MENTAL HEALTH INSIGHTS: A TEXT ANALYTICS STUDY ON UUM STUDENTS

Abstract

This study explores the application of sentiment analysis using a BERT-based model on confession texts from a university social media platform to understand student mental health at Universiti Utara Malaysia (UUM). Mental health issues among university students are often shared anonymously through platforms like UUM Confession. This research aims to identify prevailing emotional trends and sentiments expressed by students. The methodology involves data crawling from Telegram, data preprocessing, and classification into positive, negative, and neutral sentiments using a fine-tuned BERT model. The study also uses keyword extraction and visualisation techniques to support straightforward interpretation of a massive dataset. Results reveal a high occurrence of negative sentiments, often associated with stress, anxiety, and relationship issues, highlighting the urgent need for mental health interventions. The findings contribute to the broader field of educational data mining and provide actionable insights for university counsellors and administrators.