论文部分内容阅读
目的运用SVM和ARIMA方法对我国病毒性肝炎发病率进行预测,对拟合结果进行比较。为病毒性肝炎的预防提供科学依据。方法利用中国卫生统计年鉴1995-2014年的病毒性肝炎发病率数据分别建立SVM和ARIMA拟合模型,并对拟合效果进行比较。截取近年时序资料适当设置滑动窗口、映射关系和训练参数,借助MATLAB软件完成数据智能训练、仿真和预测;ARIMA法用于发病率序列拟合建模,借助SAS软件最优定阶识别、外推预测。结果病毒性肝炎发病率SVM模型和ARIMA模型SSE和MAPE分别为229、289,3.53%、3.86%。SVM模型拟合效果优于ARIMA模型,SVM模型预测2015-2017年病毒性肝炎预测发病率为(1/10万)分别为84.31、83.21、82.27。结论SVM法可用于时序建模,ARIMA法理论成熟且为经典方法。SVM模型拟合效果优于ARIMA模型,模型拟合要充分考虑数据特征。
Objective To predict the incidence of viral hepatitis in China by using SVM and ARIMA methods and compare the results of fitting. Provide a scientific basis for the prevention of viral hepatitis. Methods SVM and ARIMA fitting models were established respectively according to the data of incidence of viral hepatitis from 1995 to 2014 in China Health Statistical Yearbook, and the fitting results were compared. Intercept the sequence data in recent years, set the appropriate sliding window, mapping and training parameters, with MATLAB software to complete the data intelligent training, simulation and prediction; ARIMA method for the incidence rate of sequence fitting modeling, with SAS software optimal order identification, extrapolation prediction. Results The incidences of viral hepatitis in SVM model and ARIMA model were 229,289,3.53% and 3.86%, respectively. SVM model fitting better than the ARIMA model, SVM model predicts the 2015-17 incidence of viral hepatitis (1/100) were 84.31,83.21,82.27. Conclusion SVM method can be used for timing modeling, ARIMA method theory is mature and classical method. SVM model fitting better than ARIMA model, model fitting to fully consider the data characteristics.