中国媒介生物学及控制杂志 ›› 2021, Vol. 32 ›› Issue (4): 475-480.DOI: 10.11853/j.issn.1003.8280.2021.04.019

• 技术与方法 • 上一篇    下一篇

基于贝叶斯网络的肾综合征出血热发病率预测模型研究

李皓晨1, 齐滢滢1, 张翀1, 韩文菊1, 沈铁峰2, 李德强2, 关鹏1, 黄德生1,3   

  1. 1. 中国医科大学公共卫生学院流行病学教研室, 辽宁沈阳 110122;
    2. 葫芦岛市疾病预防控制中心, 辽宁 葫芦岛 125000;
    3. 中国医科大学公共基础学院数学教研室, 辽宁沈阳 110122
  • 收稿日期:2020-10-28 出版日期:2021-08-20 发布日期:2021-08-20
  • 通讯作者: 黄德生,E-mail:dshuang@cmu.edu.cn
  • 作者简介:李皓晨,女,在读硕士,主要从事传染病流行病学研究,E-mail:hcli@cmu.edu.cn
  • 基金资助:
    国家自然科学基金(71974199);中国医科大学新冠肺炎疫情防控相关科研攻关基金资助项目(医大科发〔2020〕12号)

A Bayesian network-based prediction model for the incidence of hemorrhagic fever with renal syndrome

LI Hao-chen1, QI Ying-ying1, ZHANG Chong1, HAN Wen-ju1, SHEN Tie-feng2, LI De-qiang2, GUAN Peng1, HUANG De-sheng1,3   

  1. 1. Departnnent of Epidemidogy, School of Public Health, China Medical University, Shenyang, Liaoning 110122, China;
    2. Huludao Center for Disease Control and Prevention, Huludao, Liaoning 125000, China;
    3. School of Fundamental Sciences, China Medical University, Shenyang, Liaoning 110122, China
  • Received:2020-10-28 Online:2021-08-20 Published:2021-08-20
  • Supported by:
    Supported by the National Natural Science Foundation of China (No. 71974199) and China Medical University Special Grant for COVID-19 Prevention and Control (No. CMU-K-2020-12)

摘要: 目的 利用贝叶斯网络研究辽宁省葫芦岛市肾综合征出血热(HFRS)的影响因素并构建发病率预测模型。方法 收集葫芦岛市2008年1-10月HFRS监测点的发病数据、宿主疫情数据及气象数据,采用禁忌搜索算法对贝叶斯网络进行结构学习,采用最大似然估计对贝叶斯网络进行参数学习。结果 葫芦岛市HFRS发病与鼠密度、当月的平均风速和日照时数、滞后1个月的平均最高气温、相对湿度和归一化植被指数、滞后2个月的平均气温、平均最低气温、平均气压和降水量在0.01水平上相关,相关系数分别为0.691、0.689、0.345、-0.635、-0.631、-0.674、-0.714、-0.746、0.650和-0.643。利用气象和宿主资料对HFRS发病率进行预测时,贝叶斯网络模型的预测准确率为85.00%(17/20),精确率为83.33%(10/12),受试者工作特征曲线下面积为0.919。结论 基于贝叶斯网络构建的发病率预测模型对葫芦岛市HFRS的预测准确率较高,对HFRS的防控有一定的参考价值。

关键词: 贝叶斯网络, 肾综合征出血热, 发病率, 预测

Abstract: Objective To study the influencing factors for hemorrhagic fever with renal syndrome (HFRS) in Huludao, Liaoning province, China and to construct an incidence prediction model using a Bayesian network. Methods Data about incidence, host, and meteorological conditions of HFRS were collected at surveillance sites in Huludao from January 2008 to October 2018. The tabu search algorithm was used to learn the structure of the Bayesian network, and maximum likelihood estimation was applied to estimate Bayesian network parameters. Results At the 0.01 level, factors associated with the incidence of HFRS in Huludao included: rodent density; the average wind speed and sunshine hours in the contemporaneous month; the average maximum temperature, relative humidity, and normalized difference vegetation index with a one-month lag; and the average temperature, average minimum temperature, average atmospheric pressure, and precipitation with a two-months lag. The correlation coefficients were 0.691, 0.689, 0.345, -0.635, -0.631, -0.674, -0.714, -0.746, 0.650, and -0.643, respectively. When using meteorological and host data to predict the incidence of HFRS, the accuracy of the Bayesian network model was 85.00% (17/20), the precision was 83.33% (10/12), and the area under the receiver operating characteristic curve was 0.919. Conclusion The Bayesian network-based incidence prediction model shows a relatively high prediction accuracy for HFRS in Huludao, providing a certain reference for HFRS prevention and control.

Key words: Bayesian network, Hemorrhagic fever with renal syndrome, Incidence, Prediction

中图分类号: