一种基于海量数据重复聚类的信用卡评级方法

CREDIT CARD RATING BASED ON REPEAT CLUSTERING IN MASSIVE DATA

对信用卡客户进行评级以提供更有针对性的服务是银行卡产业链的核心工作,对于银行以及银行卡组织有其重要的意义。目前,信用卡的评级工作多借助于对历史数据的观察和统计,人工决策参与程度高,信息利用率低下。利用信用卡交易记录得到的海量数据资源,使用基于传统方法改进的聚类算法对信用卡进行评级。改进的聚类方法通过重复迭代实现不同等级信用卡的最合适分离方案,并同时可以实现对高风险信用卡的规避,提高了信息利用率,并取得了较为理想的效果。

To rate cardholders so as to offer more specific services to them is the core work in bankcard industry chain,for banks and banknets this has the important significance. Currently the works of rating credit cards often fall back on the survey and statistics of historical data. People engage a lot in this process but use very little information of the data. We make use of the massive data gathered from credit cards transaction records,and rate the credit cards by clustering algorithm which is improved based on traditional methods. The improved clustering algorithm,through repeated iterations,realises the most appropriate separation scheme for the credit cards of different levels.Meanwhile it is able to circumvent those credit cards with high risks,and improves the information utilisation. Our work achieves desired results.