ARIMA和ARIMA-SVM组合模型在江苏省艾滋病发病预测中的应用

Application of ARIMA and ARIMA-SVM hybrid model in the prediction of HIV incidence in Jiangsu

目的建立差分自回归滑动平均求和(autoregressive integrated moving average,ARIMA)模型,在此基础之上引入支持向量机(support vector machine SVM)方法,构建ARIMA-SVM组合模型对江苏省2022年上半年艾滋病病例报告数进行预测,比较ARIMA以及ARIMA-SVM模型的的预测效果,为江苏省艾滋病防控提供科学参考。方法以江苏省2016年1月—2021年12月艾滋病病例报告数据为基线,利用R软件,构建ARIMA模型对2022年1—6月艾滋病病例报告数进行预测,并用SVM对ARIMA模型中的非线性特征进行预测,建立ARIMA-SVM模型,与实际病例报告数进行比较,评价两种模型的预测效果。结果江苏省2016年1月—2021年12月共登记艾滋病病例报告数9109例,月均发病127例。构建最佳的ARIMA模型的为ARIMA(0,0,1)(0,1,1)12,MAPE为53.27%.ARIMA-SVM模型的MAPE为14.67%,在相对误差上,除6月外,ARIMA-SVM组合模型各月的相对误差均低于ARIMA模型。结论与ARIMA模型相比,ARIMA-SVM模型的误差较小,预测效果更好。

Objective To construct a differential autoregressive integrated moving average(ARIMA)model and support vector machine(SVM)introduced ARIMA-SVM hybrid model to predict the number of AIDS cases in Jiangsu province during the first half year of 2022.The performance of ARIMA model and ARIMA-SVM hybrid model was assessed and compared.Methods ARIMA model was constructed using R software based on the data of reported AIDS cases in Jiangsu Province from January 2016 to December 2021,and SVM was introduced to predict the nonlinear features in the ARIMA model to establish an ARIMA-SVM hybrid model,both models were used to predict the number of AIDS cases during January and June 2022,the predictive efficiencies were compared.Results From January 2016 to December 2021,a total of 9,109 AIDS cases were reported in Jiangsu Province,with an average monthly cases of 127.The best constructed ARIMA model was ARIMA(0,0,1)(0,1,1)12,with MAPE of 53.27%.The MAPE of ARIMA-SVM hybrid model was 14.67%.The predictive results showed that the relative error of ARIMA-SVM hybrid model in each month(except in June)was lower than that of ARIMA model.Conclusion Compared to ARIMA model,the ARIMA-SVM hybrid model is more accurate in forecasting AIDS cases.