Mental Health with Machine Learning: A Prediction-Based Intervention Chatbot for Mental Health Conversations

Academic mental health is crucial for university students as it directly influences their academic performance, personal well-being, and future professional development. Factors such as academic stress, social isolation, financial difficulties, and family expectations can negatively impact students' mental health. Early identification of these risk factors is essential for developing effective interventions to prevent more severe mental health issues. The development of a prediction-based intervention chatbot using a machine learning approach offers a promising solution. This research utilizes a dataset derived from the DASS-21 scale, which measures levels of depression, anxiety, and stress, comprising 424 records of question-and-answer variations to train the chatbot. The dataset was divided into training and testing sets with an 80:20 ratio, allowing for unbiased model evaluation. The study compared the performance of two machine learning models: Random Forest Classifier (RFC) and Multinomial Naive Bayes (MNB). Results showed that RFC outperformed MNB, achieving an accuracy of 82.5%, along with high precision, recall, and F1-score, demonstrating its effectiveness in identifying relevant conversations and providing appropriate interventions. Although MNB is faster and more efficient, its lower accuracy of approximately 66% presents a risk of inaccurate responses. These findings highlight the potential of using RFC in a mental health intervention chatbot to offer reliable and high-quality support, making a significant contribution to both mental health and computer science by providing an effective tool for detecting and treating mental health issues among university students.