Risk Evaluation And Warning Model Of China’s Pension Fund

On the background of increasing aging population, China’s basic pension insurance fund is facing the tremendous pressures, the balance of payments is the root of healthy. Effective assessment and early warning of payments imbalances risk of basic pension insurance fund is the inevitable requirement to achieve sustainable development. This paper attempts to use the combination of rough set and BP neural network model, which using rough set to overcome the complexity of neural network, to build the combination early-warning model of payments imbalances of basic pension insurance fund.First, by collecting and collating data, we analyze the research status on pension fund risk; Second, we analyze the theoretical basis of pension funds from the aging of population and the payment of pension fund risk; third, according to the characteristics of Chinese pension insurance system, the paper compares the methods of early warning risk of pension funds; fourth, according to the system characteristics of china’s pension fund, it defines the risk of china’s pension fund, then estracts the risk factors. Based on the system characteristics and risk factors, the paper uses the application of 38 principle and factor analysis to assess China’s basic pension fund risk. Fifth, before building early warning indicator system, it tries to improve the existing system of early warning indicators from China’s national conditions, mainly to build a pension fund risk index system of "system account separation" and non-linear mesh warning pension funds index system of pension fund, then builds a suitable, reliable and effective basic pension fund risk warning indicator system; sixth, based on the pre-warning model, it makes a single BP neural network training and validation, and then uses the rough set to reduce the targets for the training and validation of the combining hybrid model of rough sets and BP neural network, then compares the two results. The study found that BP neural network loaded rough set is more stable and accurate than the single BP neural network, and the warning results is better, it can make & good characterization of support capabilities of basic pension insurance fund. Finally use the risk warning mode to predict the future pension fund risk.