Research And Application Of Stochastic Mortality Model Based On Decision Tree Method

Population mortality affects insurance liabilities,insurance product prices and social welfare programs.Research on mortality is essential in economics,demographics,and life and social insurance,and changes in population deaths have led to a decline in the regularity of global population mortality in the past.One of the main factors is therefore a hot issue that has attracted the attention of relevant scholars.Beginning with the Lee-Carter model in 1992,the stochastic mortality model has gradually become the mainstream model for predicting changes in mortality.Many different forms of stochastic mortality models have been proposed so far,because under different data structures,the optimal model is always uncertain,that is,each model has some defects.In order to optimize the model modeling effect under the mortality data in China,this paper uses a decision tree-based method to study the stochastic mortality model from the perspective of improving the fitting effect of the stochastic mortality model,which can not only improve the original model data fitting.The effect can further detect the shortcomings of the model,point out the direction of the model improvement,and achieve the goal of improving the accuracy of the result and optimizing the model selection.In addition,the article uses the improved mortality data fitted by the stochastic mortality model,combined with the decision tree method,to re-calculate the death toll of the entire population in China,aiming at a more accurate analysis of the cause of death in China.The specific work of the article is as follows:Firstly,based on the data of the full-age population mortality in China,two representative stochastic mortality models,the Lee-Cater model and the Renshaw-Haberman model,are taken as examples,and the death toll is based on the assumption of Poisson distribution.The Boosting regression tree is constructed by Boosting method to minimize the "deviation statistic" to improve the model fitting effect,analyze the lifting effect,and detect the model's insufficient performance in some aspects,and choose the direction for the improvement of the model form.Then,based on the mortality data of the post-elevation model fitting,the article continues to rely on the decision tree method to construct a framework for measuring the mortality of deaths of all age groups,and based on the results,the cause of death analysis of the whole age group.The results show that the decision tree can significantly improve the model fitting effect.For China's mortality data,the decision tree shows that the Renshaw-Haberman model has less improvement than the Lee-Carter model,which indicates that the Renshaw-Haberman model is more suitable for the modeling analysis of population mortality in China,and by adding personal characteristics and pairs.The capture of special events can continue to improve the modeling effect.The article analyzes the re-estimated data of death-cause mortality of all ages,and compares the trends of different age groups,causes of death with years and the proportion of causes of death.