Ecological Predictors and Network Patterns of Cyberbullying Victimisation Among Chinese Adolescents: Integrating Machine Learning and Network Analysis

Abstract

Cyberbullying victimisation (CBV) is an important concern for adolescent well-being and child protection, yet its predictors are often examined separately across ecological domains. Guided by an expanded social-ecological perspective, this study integrated machine learning and network analysis to examine CBV among 2006 Chinese adolescents. Twenty-two predictors spanning individual, family, peer, school and digital-behavioural factors were included, and six supervised machine learning models were compared after hyperparameter tuning. Results showed that random forest achieved the best performance on the held-out test. SHAP analysis identified traditional bullying victimisation, social media addiction, risky internet behaviours, difficulties in emotion regulation and perceived school climate as the most salient predictors; cyberbullying aggression, gaming addiction, child neglect and inferiority also ranked highly. Network analysis showed that traditional bullying victimisation had the highest betweenness centrality, whereas coping self-efficacy had the highest strength centrality. These findings position CBV within adolescents' wider developmental and safeguarding ecology rather than as a problem confined to online behaviour. Effective prevention therefore requires an ecologically informed safeguarding response that integrates early identification of offline peer victimisation and digital risk with coordinated school-based protection, family support and the strengthening of adolescents' coping resources.