Mortality Projection And Application For China Based On Neural Network Mortality Model

With the improvement of living standards and the improvement of medical and health conditions,mortality has continued to decline in China,the human life expectancy has been gradually extended,the ageing of the population is increasing,and the longevity risk brought about by the aging of the population has become a social issue that cannot be ignored.Longevity risk increases the future repayment pressure of the pension insurance system and is one of the important risks affecting the sustainable development of China's pension insurance system.Constructing an appropriate mortality model to accurately predict future population mortality is the basis for quantifying longevity risks and plays an important role in the government's development of relevant pension policies.Therefore,population mortality research is of great significance in the areas of longevity risk management and population policy formulation.In this paper,based on the neural network mortality model proposed by Hainaut(2018),considering the characteristics of China's mortality data,this paper adjusts the estimation methods and prediction methods of the model,and build a Chinese neural network mortality model.In the estimation method of the model,the paper use the denoising autoencoder to extract the potential time factor.On the one hand,the robustness of the network structure is enhanced.On the other hand,the potential time factor of the extraction is more stable,and the variation pattern and law of the data can be more accurately reflected.In the prediction method of the model,the ARIMA model is used to extrapolate the potential time factor.Using the mortality data by age and sex form 1994 to 2013 in China as the training set,the model was used to fit the training set data,and the population mortality data from 2014 to 2030 was predicted.In this paper,the real data of the mortality from 2014 to 2017 is used as a test set to evaluate the prediction effect of the model.By comparing the prediction results of the model with the prediction results of the Lee-Carter model,the prediction effect of the model is analyzed.The results show the neural network mortality model established in the paper can better predict the mortality of Chinese population,and its prediction results are better than the prediction results of the Lee-Carter model.Finally,due to the decline in mortality,the risk of mortality in China's social pension insurance system has increased.Based on the predicted age-specific mortality data,this paper constructs an actuarial model of pension personal account balance,analyzes the impact of mortality decline on pension personal accounts,and gives relevant recommendations.