人均预期寿命预测方法的比较及应用
Comparison and application of prediction methods of life expectancy
戴品远 1周晓燕 1王浩 1龚巍巍 1潘劲 1关云琦 1李娜 1钟节鸣1
作者信息
- 1. 浙江省疾病预防控制中心慢性非传染性疾病预防与控制所,浙江省杭州 310051
- 折叠
摘要
目的 比较自回归移动平均(ARIMA)模型、灰色模型GM(1,1)、Joinpoint回归模型和Lee-Carter模型对人均预期寿命预测的准确性,为预期寿命的目标值测算提供依据.方法 利用2019年全球疾病负担(GBD 2019)1990-2015年中国的分年龄组死亡率数据,采用R 4.2.0软件比较和评价4种模型预测2019年预期寿命的准确性,筛选准确性最高的方法用于分析浙江省1990-2019年数据,并预测2025年浙江省人均预期寿命.采用R 4.2.0软件计算年度变化百分比(APC)和平均年度变化百分比(AAPC)描述变化趋势.结果 Lee-Carter模型对中国2019年人均预期寿命预测值的准确性最高,均方差为0.005,均方根误差为0.072,平均绝对误差为0.063,平均绝对百分比误差为0.081.GBD数据测算得到的2019年中国人均预期寿命实际值为77.71岁,Lee-Carter模型预测为77.70岁,ARIMA模型预测为78.26岁,灰色模型GM(1,1)预测为78.42岁,Joinpoint回归模型预测为78.51岁.1990-2019年浙江省人均预期寿命呈逐年上升趋势,AAPC为0.44％,趋势有统计学意义(P＜0.01);分段APC显示,2004-2007年上升速度最快(APC=0.84％),2016-2019年上升速度最慢(APC=0.18％),趋势有统计学意义(P＜0.01).基于Lee-Carter模型预测得到2025年浙江人均预期寿命为83.35岁.结论 在时间序列较长的情况下,Lee-Carter模型对于人均预期寿命测算的准确性最高,更适用于测算目标值.
Abstract
Objection To Compare the accuracy of the autoregressive moving average(ARIMA)model,grey model(GM),joinpoint model,and Lee-Carter model for predicting average life expectancy,and provide the basis for measuring the target value of life expectancy.Methods Chinese mortality data by age groups of the Global Burden of Disease 2019(GBD 2019)from 1990 to 2015 were used,R 4.2.0 software was used to compare and evaluate the accuracy of the 4 models in predicting life expectancy in 2019,the method with the highest screening accuracy was used to analyze the data of Zhejiang province from 1990 to 2019,and the average life expectancy in Zhejiang province in 2025 was forecasted.R 4.2.0 software was used to calculate the annual percentage change(APC)and average annual Percentage change(AAPC),so as to describe the change trend.Results The Lee-Carter model had the highest accuracy in predicting average life expectancy in 2019 in China,the mean square deviation was 0.005,root mean square error was 0.072,mean absolute error was 0.063,and mean absolute percentage error was 0.081.The actual value of Chinese average life expectancy in 2019 calculated by GBD data was 77.71 years old,which predicted by Lee-Carter model,ARIMA model,GM model,and Joinpoin was 77.70,78.26,78.42 and 78.51 years old,respectively.The average life expectancy in Zhejiang Province from 1990 to 2019 showed an increasing trend year by year,AAPC was 0.44(P＜0.01),segmented APC showed the fastest rising rate(APC=0.84％)from 2004 to 2007,and the slowest rising rate(APC=0.18％)from 2016 to 2019(P＜0.01).Based on the Lee-Carter model,the average life expectancy in Zhejiang in 2025 is 83.35 years old.Conclusion In the case of a long time series,Lee-Carter model has the highest accuracy for the prediction of life expectancy and is more suitable for the measurement of target values.
关键词
预期寿命/预测/时间序列分析Key words
Life expectancy/Prediction/Time series analysis引用本文复制引用
基金项目
浙江省社科联资助项目(2022B64)
省软科学研究计划项目(2023C35007)
疾控科技英才孵育项目()
出版年
2024
NETL
NSTL
万方数据