Predicting Suicidal Thoughts in Adolescent Girls Based on Parent-Teen Conflict Using Machine Learning Algorithms

Abstract

Background: Parent-child conflict is a known risk factor for adolescent suicidal ideation but remains underexplored using predictive models. Objectives: This study aimed to predict adolescent suicidal ideation from parent-child conflict using machine learning. Methods: The study involved 442 adolescent girls from Tehran province selected via convenience sampling. Data were collected using the Beck Suicidal Ideation Scale (BSSI) and the Parent-Child Behavioral Conflict Questionnaire (CBQ). Analysis was conducted using Pearson correlation and four machine learning algorithms: Logistic regression, SVM, random forest, and XGBoost. Results: The findings revealed that certain aspects of parent-child conflict, such as incompatibility, feeling misunderstood, and shouting, were significantly associated with the severity of suicidal ideation, with correlation coefficients reaching up to 0.259. Machine learning models, particularly SVM and random forest, predicted suicidal ideation risk with an accuracy of 78.79%, while XGBoost showed a lower accuracy of 69.70%. Conclusions: This study emphasizes the role of family relationship quality in adolescent suicide prevention and supports using intelligent models for early screening.