随机死亡率模型的改进与预测
发布时间：2019-01-14 08:40
【摘要】：文章将多变点检测方法应用于人口死亡率预测,并对年龄别死亡率的偏差进行主成分提取,利用变点检测法分别估计了主要主成分得分随时间变化的最优变点个数及位置,据此对主成分得分进行分段线性回归拟合,从最后一段回归模型外推主成分得分的预测值,得到死亡率预测值;同时利用发达国家1951~2010年连续60年死亡率数据,对改进的PC模型与经典Lee-Carter进行比较研究,结果表明,改进的PC模型在死亡率预测的精度和稳定性方面均优于经典Lee-Carter模型,多变点检测方法提高了死亡率模型的预测精度。研究结果显示,基于奇异值分解的经典Lee-Carter模型中的时间因子和基于特征值分解的经典PC模型中的第一主成分得分反映出了几乎一致的死亡率变化趋势;经典PC模型中的第二主成分主要综合了队列效应对死亡率的影响。
[Abstract]:In this paper, the variable point detection method is applied to the prediction of population mortality, and the deviation of age-specific mortality is extracted by principal component extraction. The number and location of the optimal change points of the main principal component scores with time are estimated by using the change point detection method. According to this, the principal component score was fitted by piecewise linear regression, the predicted value of principal component score was extrapolated from the last stage regression model, and the mortality prediction value was obtained. At the same time, the improved PC model is compared with the classical Lee-Carter model based on the mortality data of developed countries from 1951 to 2010. The results show that the improved PC model is superior to the classical Lee-Carter model in the accuracy and stability of mortality prediction. Multivariate point detection improves the prediction accuracy of mortality model. The results show that the time factor in the classical Lee-Carter model based on singular value decomposition and the first principal component score in the classical PC model based on eigenvalue decomposition reflect the almost consistent trend of mortality. The second principal component in classical PC model mainly synthesizes the effect of queue effect on mortality.
【作者单位】： 厦门大学经济学院统计系;厦门大学经济学院;
【基金】：国家社会科学基金重大项目“大数据与统计学理论的发展研究”(编号:13&ZD148)的阶段性成果
【分类号】：C921
，
本文编号：2408529
[Abstract]:In this paper, the variable point detection method is applied to the prediction of population mortality, and the deviation of age-specific mortality is extracted by principal component extraction. The number and location of the optimal change points of the main principal component scores with time are estimated by using the change point detection method. According to this, the principal component score was fitted by piecewise linear regression, the predicted value of principal component score was extrapolated from the last stage regression model, and the mortality prediction value was obtained. At the same time, the improved PC model is compared with the classical Lee-Carter model based on the mortality data of developed countries from 1951 to 2010. The results show that the improved PC model is superior to the classical Lee-Carter model in the accuracy and stability of mortality prediction. Multivariate point detection improves the prediction accuracy of mortality model. The results show that the time factor in the classical Lee-Carter model based on singular value decomposition and the first principal component score in the classical PC model based on eigenvalue decomposition reflect the almost consistent trend of mortality. The second principal component in classical PC model mainly synthesizes the effect of queue effect on mortality.
【作者单位】： 厦门大学经济学院统计系;厦门大学经济学院;
【基金】：国家社会科学基金重大项目“大数据与统计学理论的发展研究”(编号:13&ZD148)的阶段性成果
【分类号】：C921
，
本文编号：2408529
本文链接：https://www.wllwen.com/shekelunwen/renkou/2408529.html
