Estimation And Applications Of VaR Based On EMD

Estimation of VaR (Value at Risk) is a concept proposed in recent years to measuremarket risk. Once raised, it is widely applied and become the standard measure of market riskdue to it is simple and clear and easy to use. Traditional methods to estimating VaR are mostlybased on the assumption of stationary time series and stable distributions. But according tothe market micro-structure, market is a “biased” system including noise trading andrational trading, and has strong time-varying, nonlinear and chaos features under the influenceof various factors such as trader’s psychology, international capital flows and governmentregulations etc. Therefore, distinguishing and separately discussing the different trading partsof different nature in the estimation of VaR will help further improve the accuracy of VaR.Through analyzing EMD (Empirical Mode Decomposition) method, We leads this theoryinto Financial Risk Prediction field and develop a VaR estimating Algorithm based on EMD.First, we introduce the origin, development, basic concepts and principles of decompositionabout EMD method. Then, an IMF (Intrinsic Mode Functions) restructuring method isproposedin accordance with the different features of market trading behavior at different timescales. By this method, the original time series can be divided into two sub-sequences ofdifferent nature: namely noise components and trend components with long-term memory.And according to the characteristics of different components, VaR estimation Algorithm isdeduced. Second, a series of samples is used to verify the effectiveness and robustness of theestimation. Then, it takes the Shanghai Composite Index as an example to cross compare andtest the estimation method’s feasibility in short-term and long-term period. The result showsthat the estimation of VaR by the method of this paper has higher accuracy than that bytraditional methods in different periods under the single-asset situations. Finally, this methodis extended to the estimation of portfolio risk and also gets good results. All the resultsdemonstrate that the method of this paper can effectively improve the VaR calculationaccuracy and will be a useful trial for estimation of VaR.