Abstract
INTRODUCTION: The global rise in mental health issues has become a significant public health challenge, exacerbated by the reluctance of many individuals to share their mental health concerns due to social stigma. Effective medical interventions and support systems are urgently needed. Researchers are increasingly turning to machine learning as a potential tool for diagnosing and addressing mental health conditions. OBJECTIVE: This systematic review identifies and categorizes machine-learning techniques applied to mental health detection, examines studies predicting mental health states, compiles available datasets, and analyzes the most frequently used algorithms for mental health assessment. METHODS: An extensive search was conducted across prominent databases such as Springer, ScienceDirect, IEEE, and PubMed, spanning the period from January 2015 to December 2024, using relevant keywords. Initially, 3,320 articles were selected based on their titles and abstracts. After careful examination, 35 articles met the inclusion criteria. Among the selected 35 studies, 14 leveraged data from online social networks to identify mental health issues, while 21 collected data through various manual means. These studies employed a diverse array of machine learning techniques, encompassing both supervised and unsupervised approaches. RESULTS: Machine learning exhibits promise in assisting with the diagnosis of mental health conditions and our studies show that machine learning is an effective and efficient way to detect mental health. However, further research is warranted in several key areas. Future studies should explore improved sampling methods, refine prediction algorithms, and address ethical considerations regarding using sensitive mental health data. Furthermore, incorporating image processing techniques could introduce a new dimension to this field. Collaboration with mental health specialists can augment the validity and impact of research outcomes in this critical domain. CONCLUSION: The systematic review underscores the potential of machine learning in addressing mental health issues and emphasizes the importance of ongoing research and collaboration to optimize its application in the field. It also shows that, although simpler and more interpretable models such as logistic regression are frequently used as baselines, the highest reported performances are usually achieved by more complex deep learning architectures, underscoring a central trade-off between model interpretability and predictive accuracy in this domain.