Research On Financial Risk Management Under Uncertainty Of Probabilistic Model

In the context of increasingly strengthened global economic cooperation,exchanges and integration,the economic,trade,financial and cultural ties between countries are closer,which also facilitates the globalization and diffusion of financial risks.The continuous occurrence of financial crises,especially the subprime mortgage crisis in the United States in 2008,and it soon spread to other developed countries around the world,causing significant losses to the entire world economic system.The original risk management system is facing challenges,people begin to consider new risk prediction models and recognize the importance of Knight uncertainty in risk measurement.With the establishment and improvement of the nonlinear expectation theory,more and more scholars begin to study the application of nonlinear expectation theory to develop risk management models.It is well known that insurance and risk are associated with each other,so this paper studies the application of traditional probability and statistical models to insurance risk measurement.However,traditional probability cannot deal with uncertainty well.In fact,many variables in the economic and financial field have Knigh uncertainty to a certain extent.The model under the framework of nonlinear expectation can introduce the uncertainty of mean value and fluctuation,so this paper only takes the capital market as the an example is given to illustrate the research on the model and its risk measurement under the parameter optimization based on G-expectation theory.Starting from the insurance risk management of linear probability system,on the basis of summarizing the premium and loss functions of insurance products.This paper focuses on the loss risk measurement of semi-continuous endowment life insurance products based on the assumption of uniform distribution of deaths(UDD).The result is that the loss risk of life insurance products is a more general expression.After a lapse of several years,China’s insurance business resumed in 1979 and developed rapidly.Taking the Chinese market in a typical economy as an example,this paper analyzes the relationship between insurance risk management and economic growth based on China’s gross domestic product(GDP)and insurance data,the paper mainly used vector autoregressive(VAR)model to carry out correlation statistical analysis.Furthermore,it reveals the importance of long-term co-integration and positive influence between corporate property insurance,family property insurance,credit guarantee insurance,agricultural insurance and accident insurance and economic growth,and suggests to improve insurance awareness and insurance cost.Compared with the linear probability framework,the nonlinear expectation can well describe the dynamic compatibility of financial market uncertainty,which can be more effective in risk management.Therefore,this paper aims to calculate the VaR value based on the G-normal distribution model with the framework of nonlinear expectation framework,namely GE-VaR.Firstly,based on the variance estimation method of GE-VaR model involved in the existing literature,and due to the dynamic variability of financial data and the influence of extreme value in VaR backtesting.In this paper,the original sample preprocessing method and moving window selection criteria are introduced.Secondly,the dynamic data smoothing and variance estimation adopts the scheme of the combination of the bootstrap method and the leave-one-out method.After obtaining the corresponding correction mean and variance estimation,the moving window method is used to calculate the parameters of the VaR model.Finally,this paper uses the Shenzhen Component Index data from 2015 to 2021 to conduct model empirical research,and compares the performances of GE-VaR model of the improved dynamic smooth moving window method with the benchmark VaR model(historical simulation method,exponentially weighted moving average method and AR-GARCH model)under linear probability at the 99%confidence level.The results show that:using the violation rate and deviation as the indicators of model performance analysis,the improved GE-VaR has lower deviation than the baseline VaR model,and the violation rate is closer to the significance level,and well fitted the change trend of the original data.