基于人工智能算法预测我国科技期刊编辑职业倦怠的发展态势  

Predicting the development trend of job burnout in editors of Chinese science and technology journal based on artificial intelligence algorithms

作　　者：霍荣瑞 汤春丽 李璇 陈佩琴 罗惠予 游雪梅[1] HUO Rongrui;TANG Chunli;LI Xuan;CHEN Peiqin;LUO Huiyu;YOU Xuemei(Editorial Office of Chinese Journal of Oncology Prevention and Treatmeant,Guangxi Medical University Cancer Hospital,Nanning,Guangxi 530021,China)

机构地区：[1]广西医科大学附属肿瘤医院《中国癌症防治杂志》编辑部,广西南宁530021

出　　处：《职业与健康》2024年第15期2064-2070,共7页Occupation and Health

摘　　要：目的构建和初步验证一种人工智能算法以预测我国科技期刊编辑职业倦怠的发展态势。方法基于“问卷星”采用简单随机抽样方法于2023年3月1日—5月1日对我国科技期刊编辑进行职业倦怠调查。首先采用随机森林、简单贝叶斯、K近邻法、支持向量机、人工神经网络、logistic回归和梯度增强算法7种机器学习算法构建预测模型,然后采用加权投票法基于以上7种算法构建新的融合模型,最后采用受试者工作特征(receiver operating characteristic,ROC)曲线、校准曲线和决策曲线对模型的效能进行评价,SHAP值评价预测变量的重要性,并基于R shinyAPP构建模型应用软件。结果共341名编辑的数据纳入分析,职业倦怠发生率为40.2%(137/341)。融合模型的ROC曲线下面积为0.807(95%CI:0.761~0.853)、灵敏度为0.774(95%CI:0.709~0.839)、特异度为0.772(95%CI:0.709~0.836)。校准曲线显示融合模型具有良好的校准度(Hosmer-Lemeshowχ^(2)=5.036,P>0.05)。预测特征重要性前5位依次为个人月收入水平、每周工作时间、每周加班情况、过去1年有无荣誉或奖励、职责分工是否明确。决策曲线显示,当职业倦怠预测概率在20%~90%时,采取一定的干预措施可实现获益大于代价。结论基于人工智能算法初步建立了我国科技期刊编辑职业倦怠发展态势的预测模型,可为及时采取措施减轻编辑职业倦怠提供参考。Objective To construct and preliminarily validate an artificial intelligence algorithm to predict the development trend of job burnout in editors of Chinese science and technology journal.Methods A simple random sampling method was used to investigate the job burnout in editors of Chinese science and technology journals based on"questionnaire stars"from March 1 to May 1,2023.Firstly,seven machine learning algorithms,including random forest,simple Bayes,K-nearest neighbor,support vector machine,artificial neural network,logistic regression and gradient enhancement algorithm,were used to construct the prediction model,and then a new fusion model was constructed based on the above seven algorithms by weighted voting method.Finally,receiver operating characteristic(ROC)curve,calibration curve and decision curve were used to evaluate the efficacy of the model,and SHAP value was used to evaluate the importance of predictive variables.The model application software was constructed based on R shinyAPP.Results A total of 341 editors were included in the analysis.The incidence of job burnout was 40.2%(137/341).The area under the ROC curve of the fusion model was 0.807(95%CI:0.761-0.853),the sensitivity was 0.774(95%CI:0.709-0.839),and the specificity was 0.772(95%CI:0.709-0.836).The calibration curve showed that the fusion model had good calibration(Hosmer-Lemeshowχ^(2)=5.036,P>0.05).The top five importance of prediction characteristics were personal monthly income level,weekly working hours,weekly overtime,whether there had been honors or rewards in the past year,and whether the division of responsibilities is clear.The decision curve showed that when the prediction probability of job burnout was between 20%and 90%,taking certain intervention measures can achieve maximum benefit.Conclusion A prediction model for the development trend of job burnout in Chinese science and technology journal editors has been preliminarily established based on the artificial intelligence algorithm,which has certain reference value for takin

关 键 词：科技期刊 编辑 职业倦怠 人工智能 机器学习 预测 

分 类 号：R181.31[医药卫生—流行病学]