摘要
目的构建并校验基于反向传播(BP)神经网络的登革热疫情预测模型,为登革热疫情的防控工作提供方法参考。方法基于登革热疫情资料和地理环境时空数据,分析登革热的时空分布特征及登革热病例空间自相关性,并采用Pearson相关系数对广州和佛山市(广佛)的登革热疫情及其影响因素进行相关性分析;然后利用Matlab 7.0软件完成BP神经网络预测模型的构建、训练和模拟。结果 2014年8-10月广佛地区登革热病例的空间分布中,发生本地病例数最高分别为90、386和456例/km^2,疫情空间分布主要聚集在广佛交界处;广佛地区登革热本地疫情在1 km×1 km尺度上具有极显著的空间自相关性(P=0.001,Z=134.402 5,全局Moran.s I指数=0.606 5);当月(8-10月)登革热本地病例疫情与上月(7-9月)疫情(本地病例与输入性病例)、气象(温度、湿度、降雨量)、社会(人口密度、城乡居民用地、林地、耕地)等多因素间存在不同程度的相关性;基于BP神经网络的登革热疫情预测模型的预测值与真实值相关系数为0.773,均方根误差为7.522 0。结论广佛地区的登革热疫情并非随机分布,具有明显的空间聚集性;登革热的发生受多种因素综合影响,基于BP神经网络模型可以有效地预测广佛地区登革热疫情的时空分布。
Objective The prediction model of dengue fever based on back propagation（BP） neural network was constructed and verified, which provided a reference for the prevention and control of dengue. Methods Based on the temporal and spatial data of dengue fever epidemics and geographical environment, the spatio-temporal distribution characteristics of dengue fever and the spatial autocorrelation of dengue fever cases were analyzed. Pearson＇s method was used to analyze the correlation between dengue fever and various influencing factors in Guangzhou and Foshan areas. Then,Matlab 7.0 software was used to complete BP neural network prediction model construction, training and simulation.Results From August to October 2014, the highest incidence of dengue cases in Guangzhou and Foshan area was 90, 386,456 cases/km^2, respectively, and the spatial distribution of the epidemics mainly concentrated in Guangzhou（P = 0.001,Z=134.402 5）. The global Moran＇s 1 index was 0.606 5. In the same month of dengue fever, the local epidemic situation of dengue in Guangzhou and Foshan district was significantly different. The outbreaks of the local cases were correlated to the epidemics of the previous month（July, August and September）（local cases and imported cases）, meteorological（temperature, humidity and precipitation）, and social（population density, urban and rural residential land, forest,farmland） factors. The correlation coefficient between the predicted value and the true value was 0.773 and the root mean square error was 7.522 0. Conclusion Dengue epidemics in Guangzhou and Foshan areas was not randomly distributed but obviously spatially clustered. The occurrence of dengue fever is influenced by many factors and the BP neural network model can effectively predict the temporal and spatial distribution of dengue fever in Guangzhou and Foshan areas.
作者
任红艳
吴伟
李乔玄
鲁亮
REN Hong-yan;WU Wei;LI Qiao-xuan;LU Liang(State Key Laboratory of Resources and Environmental Information System, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China;College of Geographical Sciences, Fujian Normal University College;State Key Laboratory of lnfectious Disease Prevention and Control, Natural Institute for Communicable Disease Control and Prevention, Chinese Genter for Disease Control and Prevention)
出处
《中国媒介生物学及控制杂志》
CAS
2018年第3期221-225,共5页
Chinese Journal of Vector Biology and Control
基金
国家自然科学基金(41571158)
国家重点研发计划(2016YFC1201305-03)
资源与环境信息系统国家重点实验室自主创新项目(O8R8B6A0YA)~~
关键词
登革热
空间相关
影响因素
BP神经网络模型
Dengue fever
Spatial correlation
Influencing factors
Back propagation neural network model
