基于聚类分析的申贷信用等级评价方法

 为了解决申贷信用等级评价问题,介绍了解决银行申请贷款信用等级评价中聚类分析采用的基本概念及术语,提出了2种聚类算法包括基于信贷数据的聚类算法δ-kmeans;基于高维信贷数据的聚类算法ASC,并通过实验对其性能进行比较分析,实验表明:①δ-kmeans算法在信贷风险的控制上取得较好效果;②相比传统k-means和Coweb算法,ASC算法在聚类高维信贷数据上更加有效.利用k-means算法对银行信贷数据的聚类动力学关系进行分析.最后,给出了聚类分析算法在银行信贷领域应用的的难点.

The method based on clustering analysis to evaluate the credit rating of loan application

To solve the problem of credit rating of loan application,this paper introduces the basic concepts and preliminaries of clustering analysis employed in this study to handle the problem of estimation on the loan grade of clients,and presents two kinds of clustering algorithms for solving this problem.One clustering approach is based on credit data,called δ-kmeans,and the other approach can be applied to high dimensional credit data,called ASC.Extensive experiments were conducted to compare the performance between the proposed two algorithms.The experiments show that δ-kmeans algorithm achieves better results in credit risk control,but the ASC algorithm is more effective than traditional k-means algorithms and the Coweb algorithm in clustering high dimensional credit data.In addition,this paper analyzes the dynamics of cluster analysis on bank credit data in terms of the k-means algorithm.