Bayesian Constant Mean Model State Error Variance Wt Improvement And Application In The Energy Prediction

The paper firstly introduces the Bayesian statistical methods, Bayesian forecasting thought and research at home and abroad. Furthermore, the paper introduces various different Bayesian forecasting models in details, especially Bayesian Constant Mean Model (BCM Model) which is deeply studied. However, it is not easy to get the parameters Vt、Wt of BCM Modeling in the practical application. If supposed that V is known, the parameter Wt is the only decisive factor. For a successful modeling and forecasting, it is very important to process Vt, because its value decides the stable content of the model. The paper uses a kind of new method Exponent Weighted Regression method, in which the Wt is improved to obtain Improved BCM Modeling. During the improvement the different weight coefficient for the observation value sequence was assigned as the exponent law with the time order to enhance the forecasting precision and stability of the predicted model, taking advantage of prior distributed information.In addition, using the Improved BCM Modeling does the analysis on this model for the consumed energy, in which sample data are from the related statistical data in1990-2011the National Statistical Department provides. Meanwhile, comparing to BCM Modeling, the analysis on prediction results for Improved BCM Modeling is done. As a result, the validation for improved method the paper puts forward is shown. Lastly, the energy demand in our country is predicted, using the Improved BCM Model. Meanwhile, the corresponding measures are given.