Employing decision trees to predict cyberbullying victimization among Chinese adolescents and identify subgroups and their shared characteristics

Youzhi Song,Yuan Tian,Cuiying Fan,Quan Zheng,Lu Huang,Zongkui Zhou

DOI: https://doi.org/10.1007/s12144-024-05696-0

IF: 2.8

2024-02-29

Current Psychology

Abstract:Early research has revealed distinct subgroups of cyberbullying victims. However, due to the limitations of traditional statistical methods, the characterization of features in the subgroups has been relatively limited, making it challenging to gain a relatively comprehensive understanding of different subgroup members. Decision trees and machine learning techniques offer notable advantages in addressing such issues. The primary aim of this study is to develop a high-performing classifier based on self-reported data from 814 middle school students to accurately predict cyberbullying victimization and uncover the most influential factors. On this basis, the study attempts to identify different subgroups of cyberbullying victims and their shared characteristics. The results indicated that the classification tree achieved a prediction accuracy of approximately 80% on the out-of-sample dataset. The results of permutation feature importance highlighted that the most prominent predictors were the victim's cyberbullying perpetration, followed by traditional relational victimization and depression. Further examination of the visual tree structure revealed subgroups of cyberbullying victims and their shared characteristics. The findings of this study have enriched the characteristics of different subgroups of cyberbullying victims, providing a better elucidation of the pathology of cyberbullying victimization and offering empirical evidence to facilitate the development of more targeted intervention strategies for various subgroups of cyberbullying victims.

psychology, multidisciplinary