基于ARIMA模型与GM(1,1)模型的深圳市5岁以下儿童死亡率的预测研究  

Study on prediction of under five mortality rate in Shenzhen based onARIMA model and GM (1,1) model

作　　者：王嫚 杨伟康 WANG Man;YANG Wei-kang(Department of Prevention and Health Care,Maternal and Child Health Hospital of Longhua District,Shenzhen,Guangdong 518000,China)

机构地区：[1]深圳市龙华区妇幼保健院预防保健部,广东深圳518000

出　　处：《现代预防医学》2024年第3期466-470,506,共6页Modern Preventive Medicine

摘　　要：目的 比较ARIMA模型和GM(1,1)模型在深圳市5岁以下儿童死亡率中的拟合效果,并对未来3年本市5岁以下儿童死亡率进行预测,为儿童保健工作方案的制定提供科学依据。方法 以2005—2022年深圳市5岁以下儿童死亡率为原始数据,分别构建ARIMA模型和GM(1,1)模型,进行儿童死亡率的拟合,采用平均绝对误差(MAE)和平均绝对百分比误差(MAPE)来比较模型的拟合效果。结果 2005—2022年深圳市5岁以下儿童死亡率呈下降趋势,且低于全国平均水平。构建的ARIMA(1,1,0)模型和GM(1,1)模型的MAE分别为0.30、0.24,MAPE分别为11.53%、8.73%,GM(1,1)模型的拟合效果较好。利用GM(1,1)模型预测2023—2025年深圳市5岁以下儿童死亡率分别为1.27‰、1.18‰、1.09‰。结论 GM(1,1)模型对深圳市5岁以下儿童死亡率的拟合效果优于ARIMA模型,预测未来3年本市5岁以下儿童死亡率逐年下降。Objective To compare the fitting effect of ARIMA model and grey model(GM)(1,1)in the under five mortality rate(U5MR)in Shenzhen,and to predict the U5MR in Shenzhen in the next 3 years,so as to provide scientific basis for the formulation of child health care plan.Methods Taking the U5MR in Shenzhen from 2005 to 2022 as the original data,ARIMA model and GM(1,1)model were constructed to fit the child mortality rate.The mean absolute error(MAE)and mean absolute percentage error(MAPE)were used to compare the fitting effect of the two models.Results From 2005 to 2022,the U5MR in Shenzhen showed a downward trend,lower than the national average.The MAE of the constructed ARIMA model and GM(1,1)model was 0.30 and 0.24,respectively,and the MAPE was 11.53%and 8.73%,respectively.The fitting effect of the GM(1,1)model was good.The U5MR in Shenzhen from 2023 to 2025 would be 1.27‰,1.18‰,and 1.09‰,respectively,predicted by GM(1,1)model.Conclusion The fitting effect of GM(1,1)model on U5MR in Shenzhen is better than that of ARIMA model,and it is predicted that the U5MR in Shenzhen will decrease year by year in the next 3 years.

关 键 词：ARIMA模型 GM(1 1)模型 5岁以下儿童死亡率 预测 

分 类 号：R179[医药卫生—妇幼卫生保健]