Detecting signs of depression on social media: A machine learning analysis and evaluation

Depression has become a growing concern due to its detrimental effects on both personal functioning and interpersonal relationships. In contemporary society, it is of utmost urgency to research and develop systems capable of detecting symptoms of depression on social media. Our study is not merely a survey, but a comprehensive investigation aimed at uncovering valuable insights and trends in the detection of depression on social media platforms. Our findings present a consolidated map of current methodologies, highlight key trends, and, through experimental results, provide clear performance benchmarks for both post-level and user-level techniques. By integrating insights from both the extensive literature review and practical experiments, this work clarifies existing challenges, establishes performance baselines, and proposes empirically grounded future directions to advance the development of more effective and reliable depression detection systems on social networks. This work opens a promising future for addressing the challenge of detecting depression on social media and contributes to enhancing the effectiveness of depression detection systems, ultimately aiding individuals affected by the adverse effects of depression. • Social media helps detect depressive behaviors through user posts and interactions. • FastText, Word2Vec, and BERT are popular feature extraction models. • Deep learning models like LSTM and CNN are effective in detecting depression. • Post-level analysis may not capture fluctuating moods and persistent depression. • Long-term user behavior is key to identifying trends in depressive states.