使用机器学习算法预测大学生自杀尝试风险  

Predicting Risk of Suicidal Attempt among College Students Using a Machine Learning Algorithm

作　　者：杨丽 张强[1,2,3,4] 白苏妤 周玥[1,2,3] 安莉[1,2,3] YANG Li;ZHANG Qiang;BAI Su-yu;ZHOU Yue;AN Li(School of Education,Tianjin University,Tianjin 300350,China;Suicidal Behavior Research Laboratory,Tianjin,Tianjin 300350,China;Institute of Applied Psychology,Tianjin University,Tianjin 300350,China;Shenzhen School Affiliated to Sun Yat-sen University,Shenzhen 518107,China)

机构地区：[1]天津大学教育学院,天津300350 [2]天津市自杀心理与行为研究实验室,天津300350 [3]天津大学应用心理研究所,天津300350 [4]中山大学深圳附属学校,深圳518107

出　　处：《中国临床心理学杂志》2023年第3期525-529,634,共6页Chinese Journal of Clinical Psychology

基　　金：国家社会科学基金一般项目(21BSH017);天津市教委科研计划专项任务项目(2021ZDGX03)。

摘　　要：目的:使用机器学习算法,构建大学生自杀尝试风险的预测模型。方法:选取对大学生自杀尝试风险有预测作用的33个变量,采用整群取样的方法对某高校大学生进行集体施测,得到有效问卷8992份。在Python 3.9中,构建了一个包含33个特征的支持向量机模型,应用该模型识别有自杀尝试风险的大学生,并评估模型性能。结果:构建的支持向量机模型准确率为85.77%,敏感性为73.91%,特异性为86.96%,ROC曲线下面积(AUC)为0.80。自杀行为暴露史、抑郁症、自杀态度、负性生活事件和焦虑障碍是排名前五的预测因素。结论:本研究构建的支持向量机模型可以有效识别有自杀尝试风险的大学生。Objective:Using machine learning algorithm to build a prediction model of suicidal attempt risk among college students.Methods:Participants among college students were surveyed by cluster sampling method.We selected up to 33 variables that were predictive of suicidal attempt and obtained 8992 valid questionnaires.The dataset was used to construct a support vector machine model which was used to identify college students with suicidal attempt in python 3.9.And then we evaluated the performance of the model.Results:The support vector machine identified suicidal attempt with an accuracy of 85.77%,sensitivity of 73.91%and specificity of 86.96%,area under receiver operating characteristic curve(AUC)of O.80.Exposure to suicide,depression,suicidal attitudes,negative life events and anxiety disorders were the top five predictors.Conclusion:The support vector machine model constructed in this study can effectively identify college students at risk of suicidal attempt.

关 键 词：机器学习 支持向量机 大学生 自杀尝试 

分 类 号：R395.1[哲学宗教—心理学]