The Research Of Grey Multivariate Linear Regression Analysis And Application

With the development of the scientific technology and the societal economy, the information is inflating, so the decisions that the manager needs to make are increasing. If we can exactly forecast the future's development of the things, then the predicted results could provide the theoretic basis of the decision supporting system for the managers.Multivariate linear regression analysis model is one statistic method among numerous forecasting methods, which can forecast one or many dependent variables through a set of independent variables. At present, it has extremely widespread application in the production of the industry and the agriculture and the research of the science.However, the multivariate linear regression analysis has disadvantages: (1) it cannot run after the dependent variable, the reflection of the model delays more and the response time of the system is longer; (2) it is sensitive to the abnormal data of plenty of samples. The abnormal data affects the simulation and inaccurate the results when we use the model to predict the dependent variable. However,in practice, we need not only run after the dynamic change of the dependent variable, but also avoid the influence of the abnormal data to the simulation.The grey system theory is based on the time sequence. It lays stress on the importance of the latest data in the prediction and weakens the influence of the old data. Thus the grey system can run after dependent variables, exactly predicts the dependent variables and the changing tendency of the dependent variables. In addition, the grey system is based on the accumulating characteristic of the time sequence, and avoids the influence of the abnormal data to the simulation. Basing on summarizing the shortage of grey model GM (1, N) and according to the two characteristics of the gray system, we presented a new model, i.e., gray multivariate linear regression analysis model, which combines the gray system and multivariate linear regression analysis model together. On the basis of the new model, we implement the GSAS (Grey Statistics Analysis System). We implement the multivariate linear regression analysis model, grey model GM (1, N) and new model on the platform. At the same time, we make lots of tests. The results turn out that this new model can cope with the shortages of the multivariate linear regression analysis model and the grey model GM (1, N). It not only can filter out the abnormal data which affect the simulation, but also can run after the dependent variables, therefore make the prediction more exactly.In order to check up the availability of the new model, this paper uses the new model in some applications. We can forecast the dependent variables with the new model. Through analyzing the result of the forecast, finally, we get a significant conclusion that the new model is very effective in applications and the result of the forecast can provide with the theoretic basis of the decision supporting system for the managers.