中学生校园被欺凌现状及其影响因素决策树模型的建立

Current status of middle school students bullied in school and establishment of decision tree model of its influencing factors

目的对中学生校园被欺凌现状及其影响因素进行分析,构建决策树风险预测模型。方法使用多阶段整群随机抽样法,抽取1258名中学生作为研究对象,根据在校是否发生欺凌将其分为被欺凌组和非欺凌组,收集两组相关资料,使用多因素Logistic回归筛选中学生在校被欺凌的危险因素,采用SPSS Modeler软件构建中学生在校被欺凌的决策树风险预测模型,并对其预测效能进行分析。结果1258名中学生中有325名出现被欺凌现象,其发生率为25.83%;多因素回归分析结果显示,性别、吸烟、饮酒、游戏成瘾、经常被父母打骂、学习成绩以及打架行为是中学生校园被欺凌的危险因素(P<0.05);根据危险因素构建决策树风险预测模型,模型选择了吸烟、饮酒、游戏成瘾、经常被父母打骂、学习成绩以及打架行为等6个解释变量,合计6层,23个节点,且模型选择吸烟为中学生校园被欺凌最为重要的影响因素;中学生校园被欺凌的Logistic回归模型AUC值为0.620(95%CI:0.593~0.647),决策树模型AUC值为0.658(95%CI:0.631~0.684),两种模型的Delong检验结果为Z=2.89,P=0.0039。结论性别、吸烟、饮酒、游戏成瘾、经常被父母打骂、学习成绩以及打架行为是中学生校园被欺凌的独立危险因素,根据危险因素构建的决策树模型具有较高的预测效能,为后期降低中学生校园被欺凌制定决策和规划提供了有力支持。

Objective To explore the current situation and influencing factors of middle school students bullied in school,and to establish a decision tree risk prediction model.Methods A total of 1258 middle school students were selected as research objects through multi-stage cluster random sampling method,and divided into bullied group and non-bullied group according to whether they were bullied in school.Relevant data of two groups were collected,and multi-factor logistic regression was used to screen the risk factors of middle school students being bullied in school.SPSS Modeler software was used to construct a decision tree risk prediction model for middle school students being bullied,and its prediction efficiency was analyzed.Results Among the 1258 middle school students,325 students had been bullied,with an incidence rate of 25.83%.The results of the multi-factor logistic regression analysis showed that gender,smoking,drinking,gaming addiction,frequent parental scolding,academic performance,and fighting behavior were risk factors for middle school students being bullied on campus(P<0.05).A decision tree risk prediction model was constructed based on these risk factors,with smoking,drinking,gaming addiction,frequent parental scolding,academic performance,and fighting behavior as the 6 explanatory variables,totaling 6 layers and 23 nodes.The model identified smoking as the most important influencing factor for middle school students being bullied on campus.The AUC value of the logistic regression model for middle school students being bullied on campus was 0.620(95%CI:0.593~0.647),and the AUC value of the decision tree model was 0.658(95%CI:0.631~0.684).The Delong test for the two models showed Z=2.89,P=0.0039.Conclusion Gender,smoking,drinking,gaming addiction,frequent parental scolding,academic performance,and fighting behavior are independent risk factors for middle school students being bullied on campus.The decision tree model constructed based on these risk factors has a high predictive performance,providing strong support for the formulation of decisions and plans to reduce campus bullying among middle school students in the future.