The Role of Relationship Attachment Styles on Problematic Social Media Use: A Machine Learning Analysis

This study explores the influence of four relationship attachment dimensions—secure attachment, avoidance, anger, and socio-emotional competence—on problematic social media use (PSMU) among university students. Using a quantitative design, data were collected from 227 respondents selected through proportional stratified random sampling from a population of 900 psychology students. Participants completed the Relationship Attachment Scale and the Problematic Social Media Use Scale. A machine learning analysis was employed, comparing the predictive performance of five algorithms: Linear Regression, Boosting Regression, Decision Tree Regression, K-Nearest Neighbors, and Random Forest. Results indicate that secure attachment and avoidance are the most significant predictors of PSMU. Secure attachment exhibits a strong negative association, while avoidance shows a positive effect. Anger contributes modestly, and socio-emotional competence does not significantly predict PSMU. These findings suggest that promoting secure attachment may reduce problematic social media engagement, offering valuable insights for prevention and intervention strategies in educational and clinical contexts.