基于银行流水数据的个人客户信用风险评估
Personal Customer Credit Risk Assessment Based on Bank Flow Data

对客户进行信用风险评级，是当前几乎所有银行降低信贷风险的方法。现行的客户风险评级方法大多是根据客户个人信息、历史信用信息以及财务状况等数据展开的，而这些数据结构不统一，形式复杂，无法进行批处理，增加了银行的审核时间，容易导致客户的流失。但客户的流水数据结构统一，形式相对简单，可以用批处理的方式提取包含在其中的信息，并在此基础上进行初步的个人信用风险评级，为进一步精确评级做准备。根据重庆市某商业银行个人客户的流水数据，构建了银行个人客户信用风险评价指标体系，通过因子分析法对指标进行降维，建立了新的个人风险评价指标体系，然后运用熵值法对评价指标进行赋权，对各个评价指标进行加权求和得到综合风险值，最后根据综合风险值对银行客户进行了分类。依据综合风险的大小，可以将客户分为优质客户、普通客户和不良客户三类，而三类客户对银行产生的利润比率符合帕累托80/20法则。
Credit risk rating of customers is the current method of almost all banks to reduce credit risk. Most of the current customer risk rating methods are based on the personal information, historical credit information and financial status of bank customers. However, these data are not only disunion in structure, but also complex in form. Therefore, they cannot be processed in batches, which increas-es the audit time of banks and easily leads to the loss of customers. The flow data of customers are not only unified in structure, but also relatively simple in form, so the information contained in it can be extracted by batch processing, and the preliminary personal credit risk rating can be carried out on this basis to prepare for further accurate rating. According to the flow data of individual cus-tomers of a commercial bank in Chongqing, an evaluation index system of credit risk of bank indi-vidual customers is constructed. The factor analysis is used to optimize and reduce the dimensions of the index system, and then to establish a new evaluation index system of personal credit risk. Subsequently, using the entropy method to obtain the weight of each evaluation index, and the comprehensive evaluation value is obtained by weighted sum. Finally, the bank customers are clas-sified according to the comprehensive evaluation value. According to the level of comprehensive risk, the bank customers can be divided into good customers, ordinary customers and bad custom-ers, and the profit ratios generated by the three categories of customers to the bank conforms to the Pareto 80/20 rule.