KMAN Publication Inc. (KMANPUB) Journal of Adolescent and Youth Psychological Studies (JAYPS) 2981-2526 7 Serial Number 47 2026 07 10 Classifying Nonsuicidal Self-Injury Risk Based on Childhood Trauma, Emotion Dysregulation, Peer Rejection, and Depressive Symptoms Using a Support Vector Machine Classifying Nonsuicidal Self-Injury Risk Based on Childhood Trauma, Emotion Dysregulation, Peer Rejection, and Depressive Symptoms Using a Support Vector Machine 1 11 10.61838/ EN Journal Article 2025 12 16 <table> <tbody> <tr> <td> <p><strong>Objective:</strong> This study aimed to classify past-year nonsuicidal self-injury risk among Canadian adolescents using childhood trauma, emotion dysregulation, peer rejection, and depressive symptoms within a support vector machine framework.</p> <p><strong>Methods and Materials:</strong> A cross-sectional predictive classification design was conducted with 684 adolescents aged 12–17 years recruited from secondary schools in Ontario, Quebec, British Columbia, and Alberta. Nonsuicidal self-injury was assessed using the Inventory of Statements About Self-Injury, while childhood trauma, emotion dysregulation, peer rejection, and depressive symptoms were measured using validated self-report instruments. The dataset was divided into an 80% development set and a 20% independent test set using stratified sampling. Linear and radial basis function support vector machines were compared with penalized logistic regression. Hyperparameters were optimized through stratified 10-fold cross-validation, and model performance was evaluated using sensitivity, specificity, balanced accuracy, F1 score, area under the receiver operating characteristic curve, and area under the precision–recall curve. Permutation importance was used to estimate predictor contributions.</p> <p><strong>Findings:</strong> Adolescents with nonsuicidal self-injury scored significantly higher on all predictor variables than those without such behavior, with the largest effects observed for depressive symptoms, impulse-control difficulties, limited access to regulation strategies, peer rejection, and emotional abuse (all p &lt; .001). The radial basis function support vector machine outperformed the linear model and penalized logistic regression, achieving an accuracy and balanced accuracy of .854, sensitivity of .854, specificity of .854, F1 score of .778, test AUROC of .916, and AUPRC of .836. Depressive symptoms were the strongest predictor, followed by impulse-control difficulties, limited regulation strategies, peer rejection, emotional abuse, and nonacceptance of emotional responses.</p> <p><strong>Conclusion:</strong> Nonsuicidal self-injury risk was classified with high accuracy using a nonlinear combination of depressive, emotional-regulatory, interpersonal, and trauma-related variables, supporting the use of support vector machines as adjunctive tools for adolescent risk screening.</p> </td> </tr> </tbody> </table> https://www.journals.kmanpub.com/index.php/jayps/article/download/5668/11088