Abstract

This study explores data science and machine learning techniques to predict suicidal ideation in young individuals, aiming to identify the most effective subset of questions from a comprehensive questionnaire. We benchmarked various machine learning models, including ElasticNet, Ridge Regression, Support Vector Regressor (SVR), Gradient Boosting Regressor, Random Forest Regressor, and XGBoost Regressor. Ridge Regression emerged as the most suitable model, achieving a Mean Squared Error (MSE) of 146.77, Mean Absolute Error (MAE) of 8.68, and an R2 of 0.57. Utilizing SHAP (SHapley Additive exPlanations) analysis, we identified the 20 most influential questions from the dataset. This refined approach not only enhances the efficiency of the questionnaire but also aids targeted intervention strategies by facilitating the early detection of suicidal ideation. The findings provide mental health professionals with a streamlined and effective tool to assess and address suicidal tendencies among students and young people, thereby improving preventative measures and outcomes. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.