Machine learning for suicidal ideation identification: A systematic literature review

Suicide causes approximately one death every 40 s. Suicidal ideation is the first stage in the risk scale, being a potential gate for suicide prevention. Machine learning emerged as a promising tool for helping in preventing suicide through the identification of individuals at risk. Therefore, this paper presents a systematic literature review aiming to answer how machine learning can help in suicidal ideation identification. This study addresses the state-of-the-art for this research field by filtering 4,002 articles from eleven databases published up to February 2021. We analyzed the 54 filtered articles to explore twelve research questions, addressing techniques, data, devices, explainability, and additional resources. We propose a taxonomy of machine learning techniques explored in this area and a taxonomy for highlighting the current research challenges. This review found a growing interest in suicidal ideation in the last few years. In a general way, studies explored data from social media and performed a text analysis to investigate suicidal tendencies in the individuals' language. Moreover, deep learning models seem to be a tendency in this area nowadays. Future studies in suicidal ideation should investigate generic and proactive models that do not depend on users' self-report.