A Machine Learning Model to Predict Suicidal Thoughts among Adolescent Girls with Access to Social Media. A Review of Literature

Abstract

Suicidal thoughts is one of the leading factors that cause deaths among the adolescents and young adults. Suicidal thoughts have been ranked as the major cause of deaths among adolescents in Kenya. This paper presents a systematic review of literature on prediction of suicidal thoughts among adolescent girls with access to social media. The study adopted the snowballing methodology to review the relevant literature. This involved identifying relevant and current literature on modeling of suicidal thoughts. The initial set of relevant literature was obtained by searching using keywords such as social media, suicide, mental health, adolescents, self-esteem and algorithms. The process of conducting backward and forward snowballing which entail reference tracking and citation tracking respectively followed this. Boolean operators were used to narrow down the search to at least fifty research papers relevant to the topic of study. The databases that were used to search for the literature included Google scholar, Medline, TUM university catalogue, and Project MUSE. Findings from the literature review indicated that machine learning modelling could be used to predict suicidal thoughts. The results also showed that logistic regression, decision tree, AdaBoost, artificial neural network and random forest were the commonly used algorithms in predicting suicidal thoughts among adults and youths. AdaBoost had the highest prediction accuracy of 93%. However, most studies reviewed did not mention about adolescent girls in their research thus this research paper dwelt on adolescent girls to establish how to curb suicidal thoughts among that gender.