MENTAL HEALTH SIGNAL DETECTION FROM SOCIAL MEDIA POSTS

Social media platforms have become important environments where individuals openly express emotions, stress, and psychological struggles, often before seeking professional support. This study investigates the use of interpretable text analytics techniques for detecting mental health signals, particularly depression and anxiety, from social media posts using SAS Viya Text Miner. A total of 7,731 anonymized social media posts obtained from publicly available datasets were analyzed through a transparent analytical framework consisting of text preprocessing, text parsing, concept extraction, topic modeling, and rule-based categorization. The findings revealed that depression and anxiety were the most dominant mental health indicators across the dataset. Many posts reflected persistent emotional distress rather than temporary emotional fluctuations. Anxiety-related language frequently involved panic, fear, and repetitive negative thinking, whereas depressive expressions commonly conveyed hopelessness, emotional exhaustion, and social withdrawal. Secondary indicators such as sleep disturbances and unmet emotional needs also emerged as meaningful patterns. In addition, a smaller subset of posts contained severe distress signals, including suicidal ideation. The results demonstrate that interpretable text mining approaches can effectively identify meaningful psychological patterns from large-scale social media data while maintaining transparency and explainability. Although the proposed framework is not intended for clinical diagnosis, it can support mental health research, population-level monitoring, and early identification of high-risk content within healthcare and public health systems.