摘要
当前我国人口老龄化进程加快,对未来人口结构变动的分析和预测成为国家和社会关注的重大课题。构建科学的死亡率预测模型以精确预测未来的死亡率变动,是积极应对我国人口老龄化问题的重要工作。本文采用我国1994—2023年的死亡率数据,构建集成模型以改进传统Lee-Carter模型中时间因子的预测效果。本文将死亡率数据按照时间顺序分为训练集、验证集和测试集。训练集用于估计传统Lee-Carter模型的参数,验证集用于确定集成模型中子模型的个数,而测试集用于评估不同模型的预测能力。实证结果表明,集成模型能够捕捉时间因子的时序相关性,相较于Lee-Carter模型具有更高的预测精度,有效克服了Lee-Carter模型在时间因子预测上的局限性;集成模型在不同的修匀死亡率数据集和参数估计方法下均能够得到更优的预测效果,具有一定的适用性和稳健性;Lee-Carter模型高估了未来的死亡率水平,而基于集成模型得到的长寿风险度量值更加稳健。本文所提出的集成模型对于预测未来死亡率的变动情况和应对我国人口老龄化问题具有一定的实际意义。
The aging process of China population is accelerating,and the analysis and prediction of future changes in population structure is an important topic of national and societal concern.To actively address the issue of China’s aging population,it is crucial to develop a scientific mortality prediction model that can correctly predict future changes in mortality.In order to enhance the prediction of the time factor in the Lee-Carter model,this paper proposes an ensemble model using mortality data of China from 1994 to 2023.We split the mortality data into training set,validation set,and testing set in chronological order.The training set is used to estimate the parameters of the Lee-Carter model,the validation set is used to decide the number of models in the ensemble model,and the testing set is used to assess the model’s performance.Empirical results show that the ensemble model effectively captures the temporal correlation of the time factor,offering higher predictive accuracy compared to the Lee-Carter model and overcoming its limitations in forecasting the time factor.Moreover,the ensemble model consistently yields better prediction results across various graduated mortality datasets and parameter estimation methods,which validates its applicability and robustness.While the Lee-Carter model overestimates future mortality rates,the longevity risk measures derived from the ensemble model are more stable.The ensemble model proposed in this paper has practical significance for predicting future mortality changes and addressing the aging population problem in China.
作者
张连增
李浩男
Zhang Lianzeng;Li Haonan
出处
《统计研究》
北大核心
2025年第11期141-151,共11页
Statistical Research
