Research On Model For Combination Forecasting Based On Second-order Forecast Effective Measure

Combination forecasting is the approach that combined several different forecasting methods through the appropriate weighted averaging methods, so that the effective information which is provided by the various predictive methods can be utilized comprehensively. The basic principle of combination forecasting is to combine every single predictive value and different weights into a single predictive value, then the error between forecasting and actual would be reduced,so that the predictive accuracy can be improved more effective.At present, most of linear combination forecasting is based on the predictive criteria that the minimum variance method or linear regression method to establish. But in fact, such guidelines and assumptions could not well reflect the effectiveness of the prediction method. Because the different indicators series have different dimensions, so the minimum variance method or linear regression method also do not have the same dimension. In this regard, this paper set up the combination forecasting model by using the mean of prediction accuracy and the variance which can reflect the discrete degree. The main contents are arranged as follows:Chapter Ⅰ:The current domestic and international research status of combination forecasting and the criteria used by combination forecasting are described briefly.Chapter Ⅱ:Preliminary knowledge. In this part, the concept and the related mathematical operation rules of interval number are illustrated, the concept and related properties of induced ordered weighted average operator are introduced.Chapter Ⅲ:A combination forecasting model with geometric mean is established, which is based on second-order forecast effective measure, then the related properties of the model is discussed. Some concepts such as superior combination forecasting, redundant forecasting and so on are introduced. Finally, a numerical example is given to analyze the effectiveness of the proposed combination forecast model.Chapter Ⅳ:Combined second-order forecast effective measure and induced ordered weighted average operator, the combination forecasting with induced ordered weighted average operator based on second-order forecast effective measure is established. Finally, an example is proposed to verify the validity of the model.Chapter Ⅴ:The numerical prediction is extended to interval number prediction, interval combination forecasting model based on second-order forecast effective measure is established, and the existence conditions of superior combination forecasting and its effectiveness are discussed.Chapter Ⅵ:On the work of this paper is summarized and the prospect of the combined forecasting for the future work is prospected.