Research And Application On Data Mining Classification Arithmetic Based On Multiple Classifiers Fusion

Data mining classification arithmetic is an important problem in data mining research, and it has been widely used in business area and so on. Student credit classification is a typical data mining classification problem, and a new research subject in credit classification. The research on classification arithmetic based on the classifier by multiple classifiers fusion, avoiding the unilateralism for single classifier and improving its performance, is popular.This paper stems from the ministry of education Chunhui project,"Research and application on Chongqing college student personal credit system", and the Chongqing NSF project,"Research on universal data mining mode for mixed data type". This paper does research on data mining classification based on multiple classifiers fusion method. And this method is applied on college student personal credit classification. The main work and achievements of this paper are summarized as follows:Firstly, it analyzes the basic theory of data mining, classifier and multiple classifiers fusion. This is the base for research on classification arithmetic based on multiple classifiers fusion, and its application on student credit classification.Secondly, the classification arithmetic is designed, by BP neural network fusion based on AdaBoost. The primary parts and conformation are discussed. Three key problems such as adjusting the weight of the sample, training the neural network basic classifier and setting the weight of the basic classifier, are also discussed.Finally, based on the fact of the project, the classification arithmetic based on multiple classifiers fusion is applied to student credit classification. The student credit classification data items are selected concern with student loan and pretreated. The student credit classification model based on multiple classifiers fusion is designed. And this model is test and proved that it can improve the precision and extend ability of single classifier. And it is proved to be availability and effective in student credit classification. The student credit classification module is implemented.The achievements of this paper are useful for student credit classification and student loan judgment, and for the referenced parts of credit classification.