Detection and Prediction of Future Mental Disorders from Social Media Using Machine Learning, Ensemble Learning, and Large Language Models

This paper suggests a scalable and open-ended model of identifying and forecasting mental disorders based on social media data which is made up of machine learning, ensemble learning and large language models. The proposed system overcomes multiple limitations of the current research, such as a lack of focus on particular disorders, small samples, and the lack of diversity of models, by utilizing major datasets of Reddit and screening four major mental disorders at the same time, namely ADHD, anxiety, bipolar disorder, and depression.   A considerable amount of experimentation proves that large language models, especially RoBERTa and OpenAI GPT, are continuously better than classical machine-learning and ensemble approaches in clinical and non-clinical settings due to their better understanding of context. Ensemble learning methods make robustness to operate in noisy environments but classical models offer a robust performance benchmark. The proposed framework makes the task of mental disorders prediction a challenging one, yet the results of the method are promising as they exceed the previously reported standards and prove the possible nature of early risk assessment through linguistic patterns based on social media usage.   On the whole, the results indicate that ensemble strategies can be used in conjunction with transformer-based language models to perform large-scale mental-health analysis. The suggested system has a high potential of aiding in the early intervention, proactive monitoring, and decision support to healthcare professionals. Such data-driven systems with additional refinement and ethical implementation have the potential to produce a positive contribution to enhancing both the accessibility and timeliness of mental-health care.

[1] Wagay, F. A., et al. “An Efficient Approach to Detecting Mental Illness on Online Social Media.” Artif. Med. (2026). [2]   Aragón, M. E., et al. “Adapting Language Models for Mental Health Analysis on Social Media.” Artif. Med. (2025). [3]   Kerasiotis, M., et al. “Depression Detection in Social Media Posts Using Transformer-Based Models.” Soc. Netw. Anal. Min. (2024). [4]   Hameed, S., et al. “Explainable AI-Driven Depression Detection from Social Media.” Front. Artif. Intell. (2025). [5]    Ovcharuk, O. et al. “Detecting Mental Disorders in Social Media Using Advanced Machine Learning.” medRxiv (2025). [6]   Phiri, D., et al. “Text-Based Depression Prediction on Social Media Using Machine Learning.” J. Med. Internet Res. (2025). [7]     ResearchGate. “Detection of Mental Illness from Reddit Data using Deep Learning.” (2025). [8]      ResearchGate. “Detecting Mental Health Issues Through Social Media Using NLP & ML.” (2024). [9]     ResearchGate. “AI and NLP for Mental Health Prediction from Social Media: A Decade of Progress.” (2025). [10]   ArXiv. “Mental Multi-class Classification on Social Media: Benchmarking Transformer vs LSTM.” (2025).