Research On VaR Measurement Based On Deep Learning

Influenced by the US subprime mortgage crisis,the global economy suffered heavy losses.Although the subprime mortgage crisis has been gradually away,but it remains in the hearts of people the shadow for some time,which is still difficult to erase,so far elicits lingering fear.So people continue to reflect on the nature of its occurrence,and the concept of financial risk management is at this time gradually taken seriously.Financial market risk is caused by the uncertainty of future fluctuations in financial assets.As the volatility of financial assets will drive the volatility of its value,these fluctuations,on the one hand,create the activity and liquidity of the financial market,so that the performance of various economic assets as a value movement,but on the other hand will lead to economic transition virtual,Uncertainty will be infinitely enlarged,to the investors,businesses,society,the country brought huge losses,and even lead to financial crisis.Financial risk management is to use a variety of technical means to identify the largest possible loss of each portfolio,and on this basis for analysis and decision-making,so as to maintain the healthy and stable development of financial markets.Financial risk measurement is the core and foundation of financial risk management.It is the primary task of financial risk management,which is of great significance to financial risk management.The traditional financial risk measurement method is represented by the volatility method,and the financial risk is measured by measuring the variance or standard deviation of the return on financial assets.This method only describes the degree of deviation of financial assets returns,can not explain the direction and loss of the level of deviation,the more obvious limitations,and therefore no longer adapt to the development of the times.In 1993,VaR as a new financial risk measurement tool came out,breaking the volatility method as the representative of the traditional measurement method of dominance.Through the quantitative calculation of financial risks,risk analysis can be carried out effectively,and the risk is exposured more intuitively.It has been widely applied to the measurement of financial market risk,and has played a significant role in the quantitative management of financial risks.This makes VaR quickly become the new standard of financial market risk management,and more and more widely used in financial market risk quantitative research.Although VaR has been studied for many years.However,there are few researches on the improvement of VaR calculation methods,and most of them focus on the application of VaR in various fields.In particular,China’s research on VaR is relatively late,many of which are based on mature research results abroad.Emphasis is placed on concepts,methods and empirical studies using VaR methods.Few scholars have proposed improvements to VaR,Thus ignoring the shortcomings of VaR-based risk measurement.The VaR method is to estimate the maximum possible future losses by statistically analyzing the past earnings characteristics of the financial assets.Therefore,in the process of calculating VaR,its accuracy depends on the assumption of the distribution of return on the financial assets studied and the estimation of its variance.This means that the risk measurement method based on VaR has the problem of lack of knowledge about the characteristics of the sample data,which will lead to the inaccuracy of the risk measurement and even the large deviation.The progress of science and technology has led to the change of financial market.The rapid development and spread of artificial intelligence in the world has aroused great concern from scholars.In recent years,the use of Deep Learning processing of Big Data is set off a wave.Which provides a powerful platform to quantify the financial market risk,which lays the foundation for improving the accuracy of the risk measurement and is of great significance to the financial risk management.The stock market is the most important part of financial market,it is the barometer of macro-economy.Because the stock market is not only a place for enterprises to raise funds,it is an important place for public investment.Stock as the most popular financing tool,and people’s economic life is closely related.In addition,the nature of stock investment is risk investment,therefore,the volatility of the stock market reflects the volatility of financial markets,that is,financial market risk.In summary,the volatility of the stock market has a very significant significance.Therefore,this paper takes the stock market of our country as an example.On the basis of existing literatures,this paper puts forward the VaR measurement based on the Deep Learning for the shortcomings of the current VaR methods.First of all,the traditional sense of the loss to improve the use of expected losses,and thus more in line with the reality of the people of the loss of a variety of definitions.Secondly,we establish the ARCH family model for the stock return data and the deep artificial neural network model for the expected loss,and then make a more accurate forecast of VaR.It is found that VaR measurement under Deep Learning is more accurate than VaR measurement under ARCH family model.Which shows that the VaR measurement based on Deep Learning has better prediction effect.