Machine learning-based prediction of suicide attempts among adolescents: a national study using explainable...

Area of Science:

• Adolescent Health
• Artificial Intelligence in Healthcare
• Mental Health Research

Background:

• Adolescent suicide attempts pose a significant public health challenge.
• Accurate prediction models are crucial for early intervention and prevention strategies.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting adolescent suicide attempts.
• To identify key predictors of suicide attempts using explainable artificial intelligence (XAI).

Main Methods:

• Utilized pooled data from the Korea Youth Risk Behavior Web-Based Survey (2017-2024).
• Developed and compared logistic regression, random forest, and XGBoost models for suicide attempt classification.
• Employed SHapley Additive exPlanations (SHAP) to quantify predictor contributions.

Main Results:

• Machine learning models demonstrated moderate predictive performance for suicide attempts.
• XGBoost achieved the highest F1 score, while random forest showed the highest sensitivity under screening conditions.
• Hopelessness, school violence, and perceived stress were identified as the strongest predictors, with SHAP enhancing interpretability.

Conclusions:

• Machine learning models offer a viable approach for predicting adolescent suicide attempts.
• Psychological and social factors are critical predictors, highlighting the need for holistic assessment.
• Explainable AI (XAI) methods like SHAP improve model transparency and support potential population-level screening applications.