Study On Prediction Ability Of Nerual Network Quantile Regression Model

Neural network quantile regression model (QRNN) is a combination of neural network and quantile regression, it can fully exert the advantage of neural network and quantile regression. On the one hand, it can describe the nonlinear structure of financial system through neural network, on the other hand, it may reveal the conditions distribution of the response variable by quantile regression. Therefore, neural network quantile regression model can fully promote the function of quantile regression and always get the desired empirical results in practice.This paper applied neural network quantile regression model to research of financial system forecasting, one is VaR risk measure of stock market, the other is conditional density forecasting of RMB exchange rate.In the aspect of VaR risk measure based on neural network quantile regression model, we combined neural network quantile regression model with POT method from extreme value theory to solve the problem of extreme VaR risk measure. It is difficult to choose an appropriate nonlinear functional form in nonlinear quantile regression and give an accurate measure of extreme VaR due to the data sparsity in the upper or lower tail of financial time series distribution. To this end, we proposed to describe the nonlinear structure in financial system through neural network quantile regression model (QRNN). Furthermore, we improved the ability to handle tail data information for nonlinear quantile regression using the POT method from extreme value theory, and advanced a new financial risk measure method:QRNN+POT. We studied in detail how to implement the new method and provided the procedure of the new method to estimate extreme VaR. For empirical illustration, we selected four worldwide stock markets to test the performance of the new method and classical nonlinear quantile regression models both in-sample and out-of-sample.In the aspect of RMB exchange rate forecasting based on neural network quantile regression model, we gave conditional probability density forecasting method of RMB exchange rate. We took actual RMB exchange rate as output variable and its influencing factors as input variable, built neural network quantile regression model to estimate conditional quantile of RMB exchange rate; and then realized the complete probability distribution forecasting of RMB exchange rate through probability density forecasting method. We selected RMB exchange rate against US dollar for empirical illustration, we compared the performance of probability density forecasting for the QRNN model with those for the linear quantile regression, BP neural network and linear mean regression model.The empirical results show that:first, the classical nonlinear quantile regression models perform bad in extreme VaR measure although they have good performance in normal VaR measure, QRNN+POT method however turns out to be superior to those classical models in terms of the accuracy of extreme VaR measure and is able to describe the extreme risk effectively during the financial crisis; second, neural network quantile regression model significantly improve the accuracy of RMB exchange rate forecasting, the probability density forecasting results of neural network quantile regression model can provide more abundant information for the future trend of RMB exchange rate, and also provide convenience for scientific decision-making.