由于受信用卡业务的高额利润和巨大市场空间的引诱，各大银行都开始加入或增加了在信用卡业务的投入，将其作为重点业务加以发展。但往往存在盲目发卡未合理的选择客户、风险管理等问题。本文通过回归分析，建立模型，根据客户信息预测客户的信用卡余额，预估资金成本。并从持卡人的收入、刷卡消费及个人信用三方面利用聚类分析方法对客户进行分类；利用持卡人的背景信息变量建立预测模型，对新客户所属类别进 行预测；分析得到优质客户与风险客户的一般特征，以便制定正确的营销策略或做好事前风险控制。

Due to the lure of high profits and huge credit market space of credit and business, the major banks are starting to add or increase the investment in the credit card business and develop it as the key business. However, there is lways existing some problems, such as issuing credit cards blindly, choosing customers unreasonably and venture management and so on. In this paper, through analysis of regression and building models, the author analyzes that we can predict card balance and estimate the cost of capital according to the information of customers. We also can classify customers according to the analysis of clustering aspects: the cardholder’s income, the spending of credit card and personal credit. Then by using the background information variables of cardholders to build predictive models, we can predict the category of new customers and gain the general characteristics of quality customers and risk customers. So according to that, we can draw up the right marketing strategies or do well in prior risk control.