Quantifying Suicidal Ideation on Social Media using Machine Learning: A Critical Review

Abstract

Suicidal ideation is one of the severe mental health issues and a serious social problem faced by our society. This problem has been usually dealt with through the psychological point of view, using clinical face to face settings. There are various risk factors associated with suicides, including social isolation, anxiety, depression, etc., that decrease the threshold for suicide. The COVID-19 pandemic further increases social isolation, posing a great threat to the human population. Posting suicidal thoughts on social media is gaining much attention due to the social stigma associated with the mental health. Online Social Networks (OSN) are increasingly used to express the suicidal thoughts. Recently, a top Indian actor industry took the harsh step of suicide. The last Instagram posts revealed signs of depression, which if anticipated could have saved the precious life. Recent research indicated that the public information on social media provides valuable insights on detecting the users with the suicidal ideation. The motive of this study is to provide a systematic review of the work done already in the use of social media for suicide prevention and propose a novel classification approach that classifies the suicide related tweets/ posts into three levels of distress. Moreover, our proposed classification task which was implemented through various machine learning techniques revealed high accuracy in classifying the suicidal posts. Among all algorithms, the best performing algorithm was that of the decision tree, with an F1 score ranging 0.95-0.97. After thoroughly studying the work achieved by different researchers in the area of suicide prevention, our study critically analyses those works and finds various research gaps and solves some of them. We believe that our work will motivate research community to look into other gaps that will in turn help psychiatrists, psychologists, and counsellors to protect individuals suffering from suicidal ideation.

Detecting Suicidal Ideation in Social Media: An Ensemble Method Based on Feature Fusion

“…Detecting suicidal ideation through social media is challenging [5]; therefore, we reviewed studies on suicidal ideation detection in social media and documented their methods, limitations, and model performance in Table 1.…”

“…This study also provides practical implications in the following ways. First, the use of machine learning techniques to analyze social media content can be beneficial in helping physicians identify and intervene in a timely manner with potentially suicidal populations [5]. Although the method proposed in this paper only improves the performance of existing suicide prediction models, continuous improvement has a positive effect on the overall suicide rate reduction.…”

“…Suicidal ideation is the initial unattempted plan, a suicide scheme is a technical method with a clear purpose, and a suicide attempt is an attempted behavior that can lead to death, all three of which are aimed at suicide and have increasing levels of depth. Suicide prevention requires rapid identification and intervention, especially during the COVID-19 pandemic when unusual lifestyles are affecting people's moods, and many people around the world are suffering from severe depressive disorders and psychological distress, and they are more likely to develop suicidal ideation [4][5][6].…”

“…However, there is still a lack of effective methods to identify potential people with suicidal ideation as early as possible so that timely interventions can be made to prevent them from resorting to suicidal behavior [7]. In recent years, researchers have examined people's mental health problems from two main perspectives [5]. One is based on a traditional perspective that relies on clinical interactions between medical staff and patients, using traditional scales and questionnaires to assess suicidal ideation.…”

“…One is based on a traditional perspective that relies on clinical interactions between medical staff and patients, using traditional scales and questionnaires to assess suicidal ideation. However, the drawback of this approach is that people are often shy or reluctant to consult a psychologist or counselor, suffer from deliberate concealment and misreporting, and do not disclose their plans before committing suicide [5,7]. Comparatively, suicide screening techniques through accessing and analyzing social media data is a growing and emerging field [8][9][10].…”