Research And Application Of Meteorological Integrating Forecast Based On Data Mining

With the development of weather forecast technology, more and more numerical prediction products have arisen, and forecast methods begin to become various. In practical applications, for each specific problem, the results of forecast methods are often inconsistent, and therefore we don't know how to choose. So we need a good approach to integrate the forecasting results of the same element predicted by different forecast methods and get an integrating method which is superior to the forecast methods; it is the integration problem of different forecast methods.This paper analyzes the problem in the traditional weight integrating methods and its applications and proposes a new integrating method based on data mining which uses BP Artificial Neural Network to build the integrating forecast classifier, in order to integrate the forecast results of BP Artificial Neural Network, Multiple Regression, Mean Generating Function and the Optimal climate model. The integrating forecast method makes use of the training set sifted from submethods to obtain the integrating forecast classifier; this classifier gets the optimal submethod directly according to the inputs of circulation factors, then uses the optimal submethod to predict and regards the result of the optimal submethod as the result of the integrating forecast method.The experiment shows that the reliability and the accuracy of our model are better than the submethods and other integrating forecast methods, and our method solves the problem that can't change weight dynamically in different prediction in the traditional weight integrating forecast methods.The main content of this paper and the results achieved are as follows:①Studying the submethods commonly used in weather forecast: BP Artificial Neural Network, Multiple Regression, Mean Generating Function and the Optimal climate model, and making a large number of comparative experiments.②Analyzing the problem in the traditional weight integrating methods and its applications and using data mining to propose a new integrating method; designing and implementing the algorithm by BP; Through a large number of experiments, comparing with the meteorological submethods and other integrating forecast method, we have proved that the reliability and the accuracy of this algorithm have greatly improved.