Predicting suicidal thoughts in a non-clinical sample using machine learning methods

Abstract

Objective: When examining the causes of suicide – an important public health problem – various psychological, social, cultural, and biological factors come to light. Given the complex nature of suicide, machine learning techniques have recently been used in psychological and psychiatric research. Machine learning is defined as the programming of computers to improve their performance using sample data or past experience. This study aims to predict suicidal thoughts in a non-clinical sample using supervised learning classification algorithms, one of the machine learning methods. This method is based on the risk and protective factors associated with suicide. Method: The Personal Information Form, Coping Attitudes Assessment Scale, and Rosenberg Self-Esteem Scale were used as data collection tools. The study comprised 1,940 participants, with ages ranging between 18 and 30 (x=20.48, SD=2.45). Results: Using the ensemble learning model with the Hard Voting approach, the prediction rate for a “yes” answer to the question “Have you had suicidal thoughts in the past year?” was determined to be 82%. Conclusion: This study is believed to contribute to prevention efforts by addressing potential future suicidal thoughts and preventing existing suicidal thoughts from evolving into actions. This contribution considers suicide-related warning signals and associated protective and risk factors.