地理学报 ›› 2012, Vol. 67 ›› Issue (4): 435-443.doi: 10.11821/xb201204001

• 方法研究 •    下一篇

传染病多维度聚集性探测方法

廖一兰1,2, 王劲峰2, 杨维中3, 李忠杰3, 金莲梅3, 赖圣杰3, 郑晓瑛1   

  1. 1. 北京大学人口研究所, 北京 100871;
    2. 中国科学院地理科学与资源研究所资源与环境信息系统国家重点实验室, 北京 100101;
    3. 中国疾病预防与控制中心, 北京 102206
  • 收稿日期:2011-03-18 修回日期:2011-10-10 出版日期:2012-04-20 发布日期:2012-06-11
  • 通讯作者: 郑晓瑛, E-mail: xyz@pku.edu.cn; 王劲峰, E-mail: wangjf@lreis.ac.cn
  • 基金资助:
    国家科技支撑计划课题(2006BAK01A13); 国家自然科学基金项目(41101431); 中国博士后科学基金项目(201004-70604)

A Study on Detecting Multi-dimensional Clusters of Infectious Diseases

LIAO Yilan1,2, WANG Jinfeng2, YANG Weizhong3, LI Zhongjie3, JIN Lianmei3, LAI Shengjie3, ZHENG Xiaoying1   

  1. 1. Institute of Population Research, Peking University, Beijing 100871, China;
    2. Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China;
    3. Chinese center for Disease Control and Prevention, Beijing 102206, China
  • Received:2011-03-18 Revised:2011-10-10 Online:2012-04-20 Published:2012-06-11
  • Supported by:
    National Key Technology R&D Program, No.2006BAK01A13; National NaturalScience Foundation of China, No.41101431; China Postdoctoral Science Foundation, No.201004-70604

摘要: 及早发现异常健康事件的苗头是有效进行传染病早期预警的关键.现有的传染病聚集性探测仅限于时间、空间或时空维度,往往容易忽略病例个人情况从其他方面反映的信息,从而造成过度预警.论文结合蚁群聚类算法和Bayesian Gamma-Poisson 模型,提出一种全新的传染病多维度聚类探测技术.研究区麻疹爆发案例证明该技术在继承以往时空聚集性探测技术思想的基础上,考虑了病例的属性信息,能更为灵敏、准确地找出传染病聚集区域.此方法在实际工作中具有潜在的重要应用价值.

关键词: 传染病, 聚集, 蚁群聚类算法, Bayesian Gamma-Poisson模型, 空间分析

Abstract: To indentify early signs of unusual health events is critical to early warning of infectious diseases. A new method for detecting multi-dimensional clusters of infectious diseases is presented in this paper. Ant colony clustering algorithm is applied to classify the cases of specified infectious diseases according to their crowd characters; then the cases belonging to the same class in terms of the space adjacency is separated; finally, the prior information about previous diseases outbreaks in the study area is applied to test the hypothesis that there was no disease cluster at various sub-regions. The detection ability of the method shows that this method does not need to accumulate case data within a long time period to detect irregular-shaped hot spots. It is useful for introducing spatial analysis to detection of infectious disease outbreaks.

Key words: infectious diseases, cluster, ant colony clustering algorithm, Bayesian Gamma-Poisson model, spatial analysis