Research And Application Of Multi-ensemble Classifier In The Local Area’s Temperature Forecast

With the advancement of science and technology, the development of information, the promotion of Meteorological Research Technology, and the growing in the field of meteorology data with each passing day so fast. It is an important task for the meteorological research section to find the valuable information from the massive meteorological data. Weather information is closely related with people’s lives. People’s social lives and production are directly influenced by the weather. If data mining can be applied into the meteorological data, It will make full use of the available information, these information not only can improve the accuracy of weather forecast and the ability of disaster weather warning, but also can guide the local industrial and agricultural production and raise the living standards of the peoples.In data mining, classification is a very important technology. There are decision tree, Bayesian networks, support vector machine and neural network in classification technology, these are all single classifiers. In order to improve the performance of classifier, some scholars proposed the concept of ensemble learning. Ensemble learning is a method that uses a number of basic classifiers to solve one problem, so a ensemble classifier is combining a number of different classifiers (basic classifiers) through a certain method using the related technology to fuse each base classifier eventually. Therefore, Ensemble learning classifier is also called classifier ensemble classifier. Experiments show that, the performance of ensemble classifiers is better than any one of single classifier significantly.Based on analytical the characteristics of meteorological data, the present situation of meteorological data mining and common used methods for meteorological data mining. This paper used Decision tree classification and ensemble learning to construct ensemble classification then analyzed and studied the meteorological data of a local meteorological station.This paper carried out the following research:1. Designed and implemented the parallel ensemble classifier prediction model based on decision tree, the ensemble classifier and decision tree classification method were used to predict the temperature of local area, each base classifier was used to predict the temperature of the local area respectively, and the result of ensemble classifier was colligated each base classifier, finally obtained each base classifier’s collaborative forecasting.2. Based on C4.5decision tree algorithm, Bagging, AdaBoost two kinds of ensemble model were designed and implemented, the random forest model was designed based on CART decision tree at the same time.3. According to local meteorological data, the application of three combination classifiers Bagging, AdaBoost and Random Forest, local temperature prediction model was designed and implemented respectively.4. The application of a local meteorological data was verified the Bagging, AdaBoost and Random Forest three temperatures ensemble forecasting models was effective, and the prediction results of three models had carried on the detailed comparative analysis from the accuracy and performance.The results of this research provided a basis decision making for the local weather bureau and also provided the guidance effect on the local residents of their social life and industrial production.