Improvement Of The IMF Component Selection Of EMD Combination Forecast Model

Financial physics is using the theory and research of applied physics method to understand and solve the problem of finance in the interdisciplinary. N.E.Huang and his collaborators proposed the innovation of EMD model in 1998, which is the way of any complex signal can be decomposed into several limited IMF component and have certain physical meaning. On the basis of previous research, we study the EMD and the reconstruction of the ARMA model fitting in selecting the optimal combination forecast model of the IMF component problems, and we put forward the cross validation by calculating CVscore values to select the most optimal combination forecast model of the IMF component. From the angle of the evaluation model, we can solve the problems of the EMD of the IMF in the screening and multiple accumulative error term, which has provided the beneficial exploration for EMD method to extend the new theory and application in the financial sector.The article mainly completed the four aspects of research work revolving around the EMD method, and the specific contents are asfollows:(1) The characteristics of the time-series forecasting methods include ARMA model, wavelet decomposition and EMD were studied, and it indicates the advantages of combination forecast model of the EMD and ARMA model refactoring for model fitting.(2) The problem to choose the optimal weight IMF component for original signal is existing in application of EMD combination forecast method, and research works many scholars have done are introduced.(3) We define the evaluation index value called CVscore based on cross validation. The empirical analysis, the number of the IMF component can be determined to choose the optimal fitting model by using CVscore value.(4) Calculated value Cvscore was proposed to select the most optimal IMF component of the combination forecast model, and we solved the problems of the IMF in the screening and multiple accumulative error term of the EMD. The empirical results show that the improved method is effective and reliable.