基于模糊积分的客户分类不确定性优化研究

Study on optimization of uncertainty of client classification based on fuzzy integral

数据挖掘技术为高效的客户分类提供了强大的支持,然而仅依靠这门技术并不能很好地完成这项任务。因为分类方法的局限性,现实数据存在信息的不确定、不完整、先验知识缺乏,研究对象的复杂性等困难导致的分类不确定性。从这个角度出发,将模糊积分融合方法与数据挖掘技术结合来减小客户分类的不确定性,提出了一种模糊密度修正方法,它利用了训练样本先验静态信息和各分类器识别结果包含的动态信息对模糊密度进行自适应动态赋值。仿真结果表明了它的有效性。

Though data mining technique provides powerful support to highly efficient client classification,the technique can not fulfill the task well alone.Becanse of the limitation of this classifying method,the classifying uncertainty,which is resulted in the difficulties such as the uncertainty,the incompleteness and the deficiency of transcendent knowledge of the information and the complexity of the research object,exists in the real data.From this point of view,the uncertainty of client classification is decreased by combining the method of fuzzy integral and the technique of data mining and a blur density correcting method is put forward to automatically adapt and dynamically evaluate the blur density by using transcendent static information of the training prototype and dynamic information included in the identified result of various classifiers.The emulational result testifies its validity.