Fisher判别分析用于疟疾流行时间分布预测

目的 探讨Fisher判别分析用于疟疾流行时间分布预测的可行性. 方法 根据月发病数占当年发病总数的构成比,将12个月份分为疟疾高发月、中发月和低发月,采用Fisher判别分析构建气象因素判别函数,并用自身回代与交叉验证检验判别效果. 结果 怀远县疟疾集中发生于7-10月(占全年的83.31%),季节性强.之前1月的最低气温(Tmin1)、前两个月的平均降雨量(R12)和近3个月的平均最高气温(Tmax012)最终进入判别方程式.自身回代和交叉验证结果显示:建立的判别函数对低发月和高发月的判别准确率分别为96.00%和80.95%,总体判别准确牢大于85%. 结论 采用Fisher判别分析构建的判别函数,可对当地疟疾流行的时间分布作出较好预测.

The application of Fisher discriminant analysis for time prediction of malaria prevalence

Objective To explore the application of Fisher discriminant analysis for time prediction of malaria prevalenee. Methods Months were divided into 3 classes, i. e higher prevalence month, moderate prevalence month and lower prevalence month, according to the ratios of malaria eases happened in the month to that in the whole year. Then the Fisher diseriminant analysis was applied to construct function based on weather factors, the distinguishing results were evaluated using self and cross validation method. Results Malaria ea-ses were eoncentralized in the period from July to October, including 83.31% malaria cases of the whole year in Huaiyuan County. The discriminant function formed including variables such as Trnin; ( the lowest temperature of last month), R12 (the average rainfall of last two months) and Tmax012 (the average maximum temperature of recent three months) had a distinguished accuracy higher than 85%. Conclusion The diseriminant function formed with Fisher diaeriminant analysis could do good work on malaria prevalence time prediction.