Sugeno测度下的征信用户聚类研究

征信系统是典型的大数据客户系统,如何对征信用户进行合理归类,进而从类别特征中分析用户的特点,是对征信系统进行深入挖掘的焦点问题之一.在征信系统中存在着大量的频数字段,这些变量的距离不能使用欧氏距离来刻画,Sugeno积分提供了一种考虑集合序的测度框架.本研究通过使用Sugeno测度下的部分序积分,构造了两集合的Sugeno距离差,进而在此基础上,重构Kmeans动态聚类算法,通过截集阈值来控制类的合并,从而更准确地对征信用户进行归类,并初步讨论了部分类用户的画像特征.

Clustering of credit users under Sugeno measure

How to classify users' clusters, then analyze the characteristics of users, and finally make user personas, is one of the key issues to further explore the credit information system. In academic research, clustering algorithms are often implemented for user classification. The clustering algorithm is based on the sample distance measure. In the credit system, there are lots of behavior variables, such as loan times, numbers of credit cards. The difference between these variables cannot be measured by numerical distance. This research is based on this purpose. By Sugeno measure we constructed Sugeno distance and experiment it with Kmeans dynamic clustering algorithm, we used 65536 records which came from the credit system during 2004 and 2009, explored the main clusters of these user, and finally drew personas of these users. With more and more ordinal variables in the credit system, this method will be more practical and useful.