Digital Psychiatry: Utilizing AI for Mental Health Assessment on Social Platforms

This study explores the use of machine learning and deep learning techniques to classify mental health conditions from text data on social media platforms, specifically from mental health-related subreddits. Various machine learning models, including logistic regression, XGBoost, support vector machine, random forest, history gradient boosting, and deep learning model feed-forward neural networks (FFNNs), have been applied to classify mental health conditions based on text data from social media platforms. Logistic regression and FFNN showed significant potential, with logistic regression achieving an accuracy of 83.07% on the test set. The FFNN, while highly accurate during training, exhibited overfitting issues, resulting in lower validation and test performance. XGBoost and history gradient boosting also performed well, with test set accuracies of 81.47% and 80.42%, respectively. By utilizing these models, the study aims to contribute valuable insights and tools for early detection and intervention in mental health conditions. The findings underscore the potential of social media analysis as a complementary approach to traditional diagnostic methods, offering opportunities for more timely and accurate identification of mental health issues.