| This paper explores the use of machine learning algorithms to detect signs of mental illnesses, such as anxiety, depression, BPD, autism, etc., in users' social media posts. Early and accurate diagnosis of mental health disorders is crucial for timely intervention and effective treatment, potentially improving the quality of life for those affected. To solve the multi-class classification problem, both traditional machine learning techniques and deep neural networks like Transformer-based models were explored. BERT models and ensemble models obtained the best results on a dataset of posts from Reddit communities. To explain the predictions obtained by the BERT models, a local intrepretable model as well as lingustic indicators were employed. Additionally, this study introduces a mental health web application that utilizes these machine learning models to provide real-time monitoring and assessment of users' mental health based on their social media activity. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.