Application Of Parallel Computing In Financial Risk Measurement

As an new research field, Computational Finance combines modern computing technology,Mathematics and the financial theory together to solve some much more complex and challenging financialproblems effectively．Financial risk measurement is very important,VaR (Value at Risk) is a very effectivefinancial risk measurement techniques, it is mainly used to measure the market risk at first,and it thengradually becomes the core of the financial market risk measurement techniques．So, how can calculate theVaR value more accurately is have great significance in both theoretically and in real life．This paper review a number of knowledge of techniques of financial risk measurement and thesignificance of this article．Based on such knowledge, it then sums up the technology of risk managementabroad and domestic．The article introduces the fundamental concepts of Computational Finance, thehigh-performance computing and the relevant background of financial risk measurement at first；Secondly,it introduces the relevant concepts of parallel computing, analysis the latest developments inhigh-performance computing and its concrete application in the financial field and explores the prospectivedirection of the high performance computing in the financial field；Then, it analysis the current situation ofthe financial risk measurement abroad and domestic, and it introduces the relevant concepts andsignificance of the financial risk measurement, the fundamental method of the financial risk measurement,calculation principle and method of the VaR, three kinds specific calculation method of VaR, the examinemethod of the VaR and the evaluation of the VaR；Finally, the article study the Monte Carlo parallelizationissues according to the characteristics of the Monte Carlo method in the environment of parallel．This paper selects the CSI300Index as the research object, and then use the Monte Carlo methods toresearch the instance application of financial risk measurement in parallel environment, and makes somerelated analysis and evaluations of the results．It can be found by through the case study of the article︰(1) There are big difference between VaR results when using different methods to estimate the VaRvalue；(2) Parallel acceleration effect is remarkable and the scalability is good on the whole；(3) There are little effect on the estimated value of VaR when changes Monte Carlo simulation times, It can obtains good numerical results and good numerical stability when uses relatively few simulationtimes if use the quasi random number sequence；(4) There are big difference between VaR results when sets confidence level using different methodsto estimate the VaR value；...