Forward Regression Analysis And It's Combined Forecast Model's Research In Practice

In previous work we have made a research on classic linear regression analysis, then we figured some deficits, forecasting termed the'Regression Analysis'out. This paper presents a new forecasting technique, which is derived from the Forward regression analysis on analytical models. This method not only takes advantage of regression analysis dealing with the linear system, but also departs from the thermo of traditional forecasting, the time series factors and many impact factors were took into consideration at the some time, the deficiency, the independent variables must be given at time series forecasting in regression analysis model, were made up, what is more, just to extend the time series model itself but neglect some other factors when predicting, were also overcame.For the forward regression analysis model, far greater properties were discussed with some stochastic simulation experiments, and the feasibility and validity of the forward regression analysis model. In order to optimize the forecasting effect of the forward regression analysis model, and overcome the limitation of the model itself, this paper was based on summarizing and time sequence features of Grey System, the theory of Gray System was introduced in forward regression analysis model, then a new combined forecast model, the gray multivariate-forward regression linear regression analysis forecasting model, was presented in this paper. At the same time, a comparison experiment among forward regression analysis model and gray model, the gray multivariate-forward'regression linear regression analysis forecasting model, was presented. Then we found that the gray multivariate-forward regression linear regression model can track the changing situation of dependent variable, this property can help us to get more accurate forecasting values, the true variation trend better.At last, the key was that we applied the forward regression linear regression analysis forecasting model into the forecasting of electricity demand in Hunan; we can conclude that this new model is rather effective in practice. The new assembled model for forecasting can provide the policy-making theoretical supports.