基于动力学聚类技术的银行信贷风险挖掘-四川大学.PDFVIP

( ) 计算机工程与设计 Computer Engineering and Design 3478 2009,30 14 计算机工程与设计 Computer Engineering and Design 开发与应用 基于动力学聚类技术的银行信贷风险挖掘 1,2+ 2 2 3 2 1 向剑平 ， 唐常杰 ， 陈 瑜 ， 胡进军 ， 左 劼 ， 易树鸿 (1. 遵义师范学院计算机科学系，贵州遵义563000 ；2. 四川大学计算机学院，四川成都610065 ； 3. 遵义市建设银行信息部，贵州遵义563000) 摘 要：借鉴物理学中动力学原理，提出基于动力学理论的聚类参数挖掘策略，并应用于银行贷款数据风险评估。定义了聚 类动力学参数挖掘概念、g- 平均、簇的 - 相似、风险相似度等概念，提出基于聚类动力学参数挖掘的聚类策略挖掘算法CSMA ( ) clustering strategy mining algorithm ，分析了该策略在不同参数下对实验结果的影响。实验结果表明，CSMA 策略使得聚类分析 的精度提高了9%~ 13% 。 关键词：数据挖掘; 银行贷款; 聚类; 动力学参数; 风险相似度 中图法分类号：TP311 文献标识码：A 文章编号：1000-7024 (2009) 14-3478-04 Risk analysis of banker ’s credit based on dynamic clustering technic 1,2+ 2 2 3 2 1 XIANG Jian-ping , TANG Chang-j ie , CHEN Yu , HU Jing-jun , ZUO Jie , YI Shu-hong (1. Department of Computer Science, Zunyi Normal College, Zunyi 563000, China; 2. School of Computer Science, Sichuan University, Chengdu 610065, China; 3. Information Department, Construction Bank of Zunyi, Zunyi 563000, China) Abstract ：Inspired by the theory of dynamic in physics, a novel strategy of clustering based on dynamic parameter mining is proposed, and is applied to the risk evaluation system of bank loan. The main contributions include: Concept of clustering based on dynamic para- meter mining are defined, the concepts over bank loan databases, such as g-mean, the -similar of cluster, and risk similarity, a CSMA (clustering strategy mining algorithm) algorithm based on dynamic parameters mining are proposed, the effect on different para