Integrated OS-GELM Model With Forgetting Mechanism And Kalman Filtering And Research On Risk Prediction Of Yu'ebao

Market risk is a kind of very important financial risk.How to establish a market risk prediction model to make a more accurate prediction of risk has also been a hot issue in financial research.Therefore,in order to improve the market risk prediction ability of Internet fund products,enhance the risk awareness of financial institutions,regulators and investors,and provide reference and basis for them when making decisions,this paper constructs new market risk prediction models to predict the risk more accurately,and verify the validity of the model through the relevant data of Yu'ebao.The specific research work of this paper is mainly carried out from the following aspects:First,an integrated OS-GELM model with forgetting mechanism and Kalman filtering theory(EFKOS-GELM model)is constructed.Since the OS-GELM model does not consider the timeliness of the time series and cannot deal with the problem of multicollinearity when updating the output weight,this paper adds the forgetting mechanism and Kalman filtering theory to the OS-GELM model,and the integration idea is used to further improve the stability and accuracy of the proposed model.Then the risk of Yu'ebao is predicted by the model,and three GARCH-type Models(GARCH model,GJR model,EGARCH model)and OS-GELM model are constructed as the control models at the same time.The predictive analysis results of the model in Chapter 5 show that: compared with the four control models,the predictive accuracy of the EFKOS-GELM model is the highest under all quantiles,so it can be said that the model is effective and more suitable for forecasting CVa R.Second,further,considering that practical problems are usually very complicated,a single prediction model may not be able to deal with these problems well.In order to better capture the complicated characteristics of practical problems,this paper considers the combination of linear model and non-linear models to build a mixed volatility prediction model based on the EGARCH model and the EFKOS-GELM model.At the same time,considering the possible correlation between the input variables,the principal component analysis method is applied to reduce the dimension of the input data of the model.Similarly,in order to better reflect the effectiveness of the mixed model constructed in this paper,four other control models are proposed.The predictive analysis results of the model in Chapter 5show that: compared with these four control models,the mixed model performs better both in terms of volatility prediction and in the final calculation of CVa R based on volatility prediction results,so we can say that the way to build the model through a mixed way iseffective.The research work in this paper not only enriches and improves the theoretical methods of financial market risk prediction,but also provides valuable ideas and references for other scholars to study risk prediction models.Moreover,through empirical analysis of Yu'ebao's risk prediction,it also provides important practical guidance for the Yu'ebao platform financial institutions,regulators and investors.