1.0
Cite Score 2025
SJR: 0.129 / SNIP: 0.140
JMERC
Owner
Janbazan Medical and Engineering Research Center (JMERC) is the owner of IJWPH.
Volume 18, Issue 3 (2026)
3 2026, 18(3): 1001-1010 |
Back to browse issues page
Request Type:
History
Received: 2026/07/16 | Accepted: 2026/09/2 | Published: 2026/09/8
Received: 2026/07/16 | Accepted: 2026/09/2 | Published: 2026/09/8
How to cite this article
Goodarzi N, Imani M, Mohsenabadi H, Shabani M. Machine Learning Classification of Lifetime Suicide Attempts Among Iranian Military Conscripts: A Gradient Boosting Approach. 3 2026; 18 (3) :1001-1010
URL: http://ijwph.daneshafarand.org/article-3-87214-en.html
URL: http://ijwph.daneshafarand.org/article-3-87214-en.html
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Rights and permissions
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Abstract (29 Views)
Aims: This study evaluated the utility of machine learning, specifically a Gradient Boosting model (GBM), in classifying lifetime suicide attempts among Iranian military conscripts to overcome the limitations of traditional linear assessments.
Methods: In a cross-sectional study of 338 conscripts, psychological data were collected using the Beck Hopelessness Scale–Short Form, Suicide Capacity Scale–Version 3, Psychache Scale, Acquired Capability for Suicide Scale–Fearlessness About Death (ACSS-FAD), Suicidal Behaviors Questionnaire-Revised (SBQ-R), and Interpersonal Needs Questionnaire. A GBM was trained to classify participants based on lifetime suicide attempt history and benchmarked against Logistic Regression. Feature importance was interpreted using SHapley Additive exPlanations (SHAP) to ensure clinical transparency.
Findings: The GBM achieved high classification accuracy (AUC = 0.9344), outperforming the Logistic Regression model (AUC = 0.9180). SHAP analysis identified the ACSS-FAD and SBQ-R as the most significant features distinguishing attempters from non-attempters. Furthermore, intense psychological pain (Psychache) emerged as a key discriminative factor in identifying the high-risk group. The GBM effectively captured complex, non-linear risk patterns.
Conclusion: "Capacity" factors are central to the psychological profile of attempters. Integrating non-linear machine learning techniques with theory-driven tools like the ACSS-FAD provides a robust, data-driven approach for accurately distinguishing high-risk individuals and tailoring targeted military suicide prevention strategies.
Methods: In a cross-sectional study of 338 conscripts, psychological data were collected using the Beck Hopelessness Scale–Short Form, Suicide Capacity Scale–Version 3, Psychache Scale, Acquired Capability for Suicide Scale–Fearlessness About Death (ACSS-FAD), Suicidal Behaviors Questionnaire-Revised (SBQ-R), and Interpersonal Needs Questionnaire. A GBM was trained to classify participants based on lifetime suicide attempt history and benchmarked against Logistic Regression. Feature importance was interpreted using SHapley Additive exPlanations (SHAP) to ensure clinical transparency.
Findings: The GBM achieved high classification accuracy (AUC = 0.9344), outperforming the Logistic Regression model (AUC = 0.9180). SHAP analysis identified the ACSS-FAD and SBQ-R as the most significant features distinguishing attempters from non-attempters. Furthermore, intense psychological pain (Psychache) emerged as a key discriminative factor in identifying the high-risk group. The GBM effectively captured complex, non-linear risk patterns.
Conclusion: "Capacity" factors are central to the psychological profile of attempters. Integrating non-linear machine learning techniques with theory-driven tools like the ACSS-FAD provides a robust, data-driven approach for accurately distinguishing high-risk individuals and tailoring targeted military suicide prevention strategies.
Send email to the article author