The Study And Application Of Trend Prediction Based On Combination Forecast Method

So called combination forecasting model is a new forecast model that synthesizes varies of individual forecasts method with different weights. In the circumstance of so many individual forecasting model are different, it get what it need in different type of single model, gather more information and forecasting techniques, in order to reduce the prediction error and improve the effect of prediction significantly. Since it was put forward by Bates J. M. and Granger C. W. J. the first time in 1969, the combination forecasting has been made great development. Theoretical study and practical application show that the combination forecasting model has higher prediction accuracy than the single forecasting model, can enhance the stability of prediction, and can adapt to future environmental change better, thus it attract many scholars.However, the models and theories of combination forecasting are incomplete due to its short history, so it is necessary to study further. The purpose of this dissertation is to do some exploratory attempt and application research in combination forecasting method, predict the trend by the combination forecasting method based on numerical prediction, which has great significance to enrich the combination forecasting methods, develop the theoretical prediction disciplines and strengthen the scientific management.In this dissertation, at first, it gives a review of the traditional prediction methods, as well as analyses the introduction of its advantages and disadvantages, then introduce the information aggregating operators which will be used in combination forecasting, it also do some exploratory attempt, deeply explore and study Markov prediction method and one-step forward prediction based on rough sets and conversion between trending structure sequence and information system, finally, by the original combination forecasting method in the foundation of numerical prediction, we bring up a combination forecasting method based on trend prediction, and applied it to the prediction of temperatureThe main research works of the dissertation are summarized as following:1. In this dissertation, we give a method to solve the problem of tie in IOWA operator, it is different from Yager's method, it offset the defect in original method which change the aggregating data, and obtain satisfactory results in experiment.2. In the foundation of the combination forecasting of numerical prediction method, the dissertation bring up a combination forecasting method based on trend prediction, the predictive values of the individual prediction methods are all language value which express the trend. To combine the term set S, we give the formula of prediction accurately, set up the most optimal model and obtain the weight. At the same time, optimize the combination prediction results according to the shortest distance. The application of combination forecasting method in the temperature prediction can prove that the method we bring up can offset the limitations of the information source and the sensitivity of model forms in the single prediction model, improve the prediction accuracy.