The increasing use of social media platforms has created vast amounts of user-generated textual data that reflect personal emotions, thoughts, and behavioural patterns. These digital footprints provide valuable insights into an individual's psychological state and can be leveraged for early detection of mental health conditions. However, traditional mental health assessment methods rely heavily on clinical interviews and self-reported questionnaires, which may not always provide timely or scalable solutions. This study proposes an intelligent hybrid machine learning framework for early mental health risk prediction using social media text analytics. The system integrates conventional machine learning models and deep learning architectures to perform multiclass classification of mental health conditions such as anxiety, depression, stress, and other psychological states. The framework incorporates comprehensive text preprocessing techniques, including cleaning, tokenization, stop-word removal, and feature extraction using advanced vectorization methods. Multiple classifiers such as Support Vector Machines (SVM), Random Forest, Logistic Regression, XGBoost, and a hybrid BiLSTM-CNN deep learning model are implemented and evaluated. To enhance performance, the proposed system applies hyperparameter optimization and dynamic model selection strategies. Experimental results demonstrate that the hybrid framework achieves high predictive accuracy and balanced performance across precision, recall, and F1-score metrics. The system provides a scalable and automated approach for mental health analysis, offering potential support for early intervention and preventive healthcare strategies.
Rao et al. (Thu,) studied this question.