摘要
目的为病毒性肝炎流行趋势分析建立可靠的CEEMD-GWO-SVR集成预测方法。方法基于分解集成思想,首先利用CEEMD分别将甲型肝炎、乙型肝炎、戊型肝炎发病数的原始时间序列分解为一系列本模特征函数,然后利用GWO-SVR对每个本模特征函数建模,最后集成各本模特征函数的预测得到最终结果。结果CEEMD-GWO-SVR比GWO-SVR、SARIMA的预测精度提高了2%~8%,其中甲型肝炎的MAPE分别降低了5.68%和7.06%,乙型肝炎降低了2.19%和4.34%,戊型肝炎降低了6.14%和4.59%。假设检验的结果也证明了CEEMD-GWO-SVR的预测与实际序列相关性更强。结论CEEMD-GWO-SVR具有较高的预测精度,可以为病毒性肝炎流行趋势分析提供可靠的预测方法。
Objective To establish a reliable CEEMD-GWO-SVR integration prediction method for epidemic analysis of viral hepatitis.Methods Based on the decomposition-integration idea,the original time series of HAV,HBV and HEV are decomposed into a series of eigenfunctions of the model by CEEMD,and then each eigenfunction of the data is modeled by GWO-SVR,and the final prediction results are obtained by integrating the predicted results of each eigenfunction of the model.Results Compared with GWO-SVR and SARIMA,the accuracy of CEEMD-GWO-SVR is improved about 2%~8%.Specially,the MAPE for HAV decreased 5.68%and 7.06%,HBV decreased 2.19%and 4.34%,and HEV decreased 6.14%and 4.59%,respectively.The hypothesis testing results also prove that the prediction results of CEEMD-GWO-SVR model is more correlated with the actual sequence.Conclusion CEEMD-GWO-SVR model has high prediction accuracy and can provide a reliable method for the prediction of viral hepatitis epidemic trend.
作者
杨慧
魏麟
胡晓斌
朱素玲
Yang Hui;Wei Lin;Hu Xiaobin(Department of Epidemiology and Health Statistics,School of Public Health,Lanzhou University(730000),Lanzhou)
出处
《中国卫生统计》
CSCD
北大核心
2022年第6期815-818,823,共5页
Chinese Journal of Health Statistics
基金
国家社科基金重大项目(20&ZD120)。
关键词
病毒性肝炎
预测
分解集成模型
Viral hepatitis
Forecasting
Decomposition-integration model
