Research on prediction model of adolescent suicide and self-injury behavior based on machine learning algorithm

Area of Science:

• Psychiatry
• Machine Learning
• Adolescent Health

Background:

• Adolescent suicidal and self-injurious behaviors are significant public health concerns.
• Identifying risk factors and developing predictive models are crucial for prevention and intervention.

Purpose of the Study:

• To explore risk factors associated with adolescent suicidal and self-injurious behaviors.
• To construct a machine learning-based prediction model for these behaviors.

Main Methods:

• Stratified cluster sampling of 3,000 high school students in Chongqing.
• Logistic regression analysis to identify independent risk factors.
• Comparison of six machine learning algorithms (MLP, RF, KNN, SVM, LR, XGBoost) for predictive modeling.

Main Results:

• Gender, impulsivity, psychoticism, neuroticism, interpersonal sensitivity, depression, and paranoia were identified as independent risk factors.
• The logistic regression model achieved the highest sensitivity (0.9948) and specificity (0.9981).
• Random forest, multi-level perceptron, and extreme gradient boosting models showed acceptable performance.

Conclusions:

• Adolescents with impulsivity, psychoticism, neuroticism, interpersonal sensitivity, depression, and paranoia are at higher risk.
• Machine learning models effectively classify and predict adolescent suicide and self-injury risk.
• Early identification enables targeted interventions to mitigate these behaviors.