摘要
目的分析重庆市2009-2014年痢疾流行特征及其与气象因素的关系,建立痢疾发病率的预测模型,并评价拟合效果,为痢疾发病预测预警提供科学方法。方法收集重庆市2009-2014年气象资料和痢疾发病数据,利用Excel2003进行数据整理,SPSS 18.0统计软件进行气象因素与痢疾周发病数的相关分析并建立多元回归模型,利用matlab 7.0软件构建气象因素与痢疾周发病率的BP人工神经网络模型。结果 2009-2014年重庆市共报告痢疾55 580例;年平均发病率为28.0/10万,发病有明显季节性,发病高峰在5~10月份,且男女发病率差异有统计学意义(χ2=112.03,P<0.01)。病例以散居儿童、学生和农民为主;痢疾周发病率与同1周和前1周的周平均水气压、周平均气温、周平均最高气温、周平均最低气温、周平均温差、周平均日照、周平均露点温度均成正相关,与周平均气压成负相关(均有P<0.05)。利用同一周与前一周的气象因素建立的BP神经网络模型,其平均绝对误差(MAE)分别为0.087和0.071;利用同一周与前一周的气象因素建立的多元回归模型,MAE分别为0.077和0.074。结论利用前一周的气象因素建立的BP神经网络模型对痢疾发病预测具有良好的拟合效果和预测能力;影响痢疾发病的主要气象因素为周平均水气压,周平均露点温度,周平均气压,周平均温度。
Objective To analyze the incidence of dysentery in Chongqing from 2009 to 2014 and its relationship with meteorological factors so as to provide a scientific basis for effective prevention and control of dysentery.Methods The data of meteorological data and dysentery in Chongqing from 2009 to 2014 were collected and analyzed using Excel 2003. Correlations were analyzed using SPSS 18. 0. The prediction model of BP neural network was established using Matlab 6. 5. The multiple regression model was established using SPSS 18. 0.Results A total of 55 580 cases of dysentery was reported in Chongqing in 2009-2014; the average annual incidence rate was 28. 0/105; the occurrence was markedly seasonal,and the peak of occurrence was May to October; the sexual difference of incidence was statistically significant（ χ~2= 112. 03, P〈0. 01）;patients were mainly scattered children, students and peasants. The weekly incidence of dysentery was positively correlated to the weekly average vapor pressure,average temperature, average maximum temperature,average temperature difference,average sunshine duration and average dew point temperature of the same week and the previous week,and was negatively correlated to the weekly average air pressure（ P〈0. 05）.Calculated using the BP neural network model established by the meteorological factors of the same week and the previous week,the forecast accuracy index MAE were 0. 087 and 0. 071,respectively; calculated using the multiple linear regression model established by the meteorological factors of the same week and the previous week,the forecast accuracy index MAE were 0. 077 and 0. 074,respectively. Conclusion The BP neural network model established by the meteorological factors of the previous week has a good fitting effect and prediction ability for the prediction of dysentery. Weakly average dew point temperature,vapor pressure,average air pressure and minimum temperature are the leading meteorological factors influencing the incidence of dysentery.
作者
肖达勇
刘勋
廖骏
赵寒
宿昆
马颖
夏宇
熊宇
陈熙
肖邦忠
李勤
XIAO Dayong;LIU Xun;LIAO Jun;ZHAO Han;SU Kun;MA Yin;XIA Yu;XIONG Yu;CHENG Xi;XIAO Bangzhong;LI Qin(Center for Disease Control and Prevention of Chongqing, Chongqing 400042, China)
出处
《预防医学情报杂志》
CAS
2018年第6期722-727,共6页
Journal of Preventive Medicine Information
基金
重庆市医学科研项目(项目编号:2015MSXM094)
关键词
痢疾
多元线性回归模型
气象因素
BP人工神经网络模型
meteorological factors
multiple linear regression
back -propagation artificial neural network
