Prediction Of VaR Risk And Price Information In Financial Market

Financial market is a complex nonlinear system and there is much information included in financial data. Analyzing and predicting financial data using some methods is a challenging research direction, but it is much of value. Risk and price are two important aspects of financial information in that predicting risk message is helpful to protect financial market's safety and predicting price message can help investors to benefit from investment. This paper does some research in these two aspects.There are several well-known facts about financial equity return series. The first one is fat-tails of the return distribution and the second is volatility clustering. In addition, recent studies show strong evidence that lots of financial equity return series exhibit long memory behavior in volatility. Considering these three features, this article constructs a dynamic VaR risk measure based on the FIGARCH-EVT model for financial equity return series. The combination can make full use of the advantages of FIGARCH model and EVT (Extreme Value Theory) method from the following two points. On the one hand, as one of the GARCH model relatives, FIGARCH model could not only deal with the equity return's heteroscedasticity, but also recover the long memory in volatility. On the other hand, the application of EVT is effective in tracking extreme losses in the study of risk measurement. It is able to capture the fat tails of the equity return distribution that shows clear non-normal behavior.It is reasonable to use artificial neural networks method to predict equity price's trend because it is close to a non-linear process. The volatility expressed in conditional variance is a prominent feature of equity return. This paper chooses conditional variance estimated by FIGARCH model as one input variable of neural networks and then establishes a neural network based on FIGARCH model.At last, these models are applied on daily returns of composite index of Shanghai stock market for forecasting purpose. The daily closing price ranges from Jan 1, 1995 to May 17, 2007. The empirical analysis indicates that the risk measure can deal with the index return's three traits and describes its dynamic VaR risk more exactly. The risk measure proposed performs higher hit rate than the dynamic VaR model based on GARCH-N. Besides, the outcomes are more rational than the static VaR model based on EVT. The outcomes also show that the FIGARCH-ANN model performs a higher hit rate than neural networks based on GARCH and neural networks without volatility as an input variable.The VaR risk measure based on the FIGARCH-EVT model and the artificial neural networks based on the FIGARCH model are of certain value for forecasting financial messages. Our work done again proves that there is apparent long memory property in Chinese stock market volatility.