Prediction Of Volatility And Calculation Of Risk Values Of Financial Markets Based On Deep Learning

The management and control of financial risk has always been the focus of finance.With the advent of big data era and the development of Internet finance,the financial data level has reached a huge amount of units,at the same time,the complexity of data,as well as the unstructured form of data is enhanced.The big data processing work is facing great challenges,especially in volatility forecasting and risk management of complex data processing,it has become difficult for traditional economic model to present specific accurate results.This paper introduces deep learning model into financial market and combines it with traditional statistical model to calculate volatility and risk value of financial market.The first chapter of this paper introduces the background,significance and current research status of the article,and gives the main research framework of the article.The second chapter is the theoretical basis of volatility and financial risk,it mainly introduces the VaR method of extreme volatility and financial risk measurement.In chapter 3,on the basis of the characteristics of time series,the MA filtering method that can decompose time series into high volatility components and low volatility components is introduced.Then,the ARMA model is introduced,and the ARMA model is used to predict the high volatility components with the depth learning model.Then ARMA-RNN combination model and ARMA-LSTM combination model based on MA filtering method are proposed respectively for short-term and long-term prediction of volatility time series.The fourth chapter carries out an empirical study with the combined model proposed in the third chapter,selects six sets of stock data,and carries out a short-term 1 forward prediction and a long-term 3 step forward prediction for its volatility.The fourth chapter in the third chapter puts forward the combination of empirical research model,selection of four groups of stock data,the volatility respectively were 1 step forward and three steps forward prediction,the empirical results show that the combined model based on MA filtering method has good effect for the forecast of volatility.The fifth chapter first introduces the steps of calculating VaR value by general Mento Carlo simulation.After that,we have made two improvements to the Mento Carlo simulation method by analyzing the statistical characteristics of the return series.For the first time,the ARMA-RNN combination model is added to the Mento Carlo simulation method,and the VaR value is estimated by the volatility estimated by ARMA-RNN combination model.The second time,on the basis of the first improvement,used distribution instead of normal distribution in Mento Carlo simulation method to calculate VaR value.Finally,two sets of data are selected for empirical analysis.The results show that the improvement effect of Mento Carlo simulation is better.The sixth chapter is the summary of this paper.