A Systematic Review of Machine Learning Algorithms for Early Detection and Prevention of Depression

Background and Objective: The rising prevalence of depression and limited access to mental health services underscore the need for machine learning algorithms in early diagnosis and prevention. The objective of this study is to conduct a systematic review to evaluate the applications and effectiveness of these algorithms in improving mental health outcomes. Methods: A systematic review was conducted following PRISMA guidelines by searching PubMed, Scopus, Web of Science and IEEE Xplore for articles published from ۲۰۱۴ to ۲۰۲۵. After screening and applying inclusion criteria, ۳۸ high-quality studies were selected. Results: Sixty-eight percent of studies focused on depression diagnosis, ۵۵% on prevention, and ۴۰% on continuous mental health monitoring. Data sources comprised electronic questionnaires, textual analyses, and behavioral data. Deep neural networks, especially transformer models like BERT and GPT, achieved accuracy between ۸۵% and ۹۱% in predicting depressive symptoms ۱۰ to ۱۲ days before clinical diagnosis. These models effectively identified hidden behavioral patterns and enabled ongoing mental health monitoring. Conclusion: Machine learning algorithms are powerful tools for diagnosing and preventing depression; however, challenges such as data privacy and methodological standardization remain. Future studies should aim to improve accuracy and develop ethical frameworks. Keywords: Machine learning, depression, early detection, mental health, neural networks

کلیدواژه ها:

Machine learning ، depression ، early detection ، mental health ، neural networks