我国某地区痢疾发病率与气象因素的关系及其预测模型  

FORECASTING MODEL FOR INCIDENCE OF DIARRHEA AND ITS RELATIONSHIP WITH METEOROLOGICAL FACTORS

作　　者：黄成钢 金如锋[2] 邱宏[3] 黄品贤[2] 王中民[2] 周霞[4] 王国复[5] 魏建子[6] 

机构地区：[1]江西省九江市卫生防疫站业务员,九江332000 [2]上海中医药大学预防医学教研室 [3]香港中文大学公共卫生学院社区与家庭医学系 [4]山东中医药大学第二附属医院神经内科 [5]国家气象中心 [6]上海中医药大学针灸推拿学院

出　　处：《现代预防医学》2009年第7期1207-1210,共4页Modern Preventive Medicine

基　　金：上海市教委科研项目(05CZ01);上海市教委高校高水平特色发展项目(沪教委财(2005)81号)

摘　　要：[目的]建立我国某地区痢疾发病率的预测模型,并探讨各种气象因素对痢疾发病率影响的相对重要性。[方法]以2000年1月～2005年12月气象因素为输入神经元,同期痢疾月发病率为输出神经元建立BP人工神经网络模型。同时以气象因素为自变量,痢疾月发病率为应变量,建立多元线性回归模型。以上两模型分别以MIV值和标准化偏回归系数确定各气象因素的相对重要性。以2006年痢疾月发病率检验以上模型的预测效果。[结果]BP人工神经网络模型的平均误差率为17.12%;非线性相关系数为0.76。多元线性回归模型的平均误差率为10.74%;非线性相关系数为0.88。多元线性回归模型表明,影响痢疾发病率的重要气象因素为平均气压、平均最低气温、平均最高气温、平均相对湿度。BP人工神经网络模型的研究结果与其基本一致,平均气压、平均相对湿度、平均最高气温、平均气温为重要性排序前4位的气象因素。[结论]对我国某地区痢疾发病率的预测可使用以气象因素为自变量的多元线性回归模型进行预测,影响疾病发病率的主要气象因素为平均气压、平均最低气温、平均最高气温、平均相对湿度。[Objective] To develop a forecasting model for the incidence of diarrhea in a city, and analyze the relative importance of meteorological factors which would influence the incidence of diarrhea. [ Methods] Back-propagation artificial neural network was developed while the data of meteorological factors from Jan 2000 to Dee 2005 were used as input neuron, and monthly incidence of diarrhea during the same period was used as output neuron. Further more, multiple linear regression was developed with meteorological factors as independent variables and monthly incidence of diarrhea as dependent variable. The mean impact value （MIV） and standardized partial regression coefficient were used to assess the relative importance of meteorological factors for the above models respectively. The data of incidence of dian-hea in 2006 was adopted to validate the predictability of both models. [ [Results] For back-propagation articial neural network, the mean relative error was 17.2%, and the nonlinear correlation coefficient was 0.76. For multiple linear regression, the mean relative error and the nonlinear correlation coefficient was 10.74%, 0.88 respectively. According to multiple linear regression, the important meteorological factors were monthly mean air pressure, minimum temperature, maximum temperature, relative humidity. Similar to the result of multiple linear regression, the descending important meteorological factors were mean air pressure, relative humidity, maximum temperature, temperature, and so on. [ Conclusion ] The best model to forecast the incidence should be multiple linear regression with meteorological factors as independent variables. Monthly mean air pressure, minimum temperature, maximum temperature and relative humidity are the leading meteorological factors which influence the incidence of diarrhea.

关 键 词：痢疾 气象因素 多元线性回归模型 BP人工神经网络模型 

分 类 号：R516.4[医药卫生—内科学]