K均值聚类算法在银行客户细分中的研究  

Simulation Study on Commercial Bank Customer Segmentation on K-means Clustering Algorithm

作　　者：樊宁[1] 

机构地区：[1]太原理工大学轻纺工程与美术学院,山西晋中030600

出　　处：《计算机仿真》2011年第3期369-372,共4页Computer Simulation

摘　　要：研究银行客户细分问题,对客户进行分类,应针对获利最大的为识别目标。为了减少主观性分析,采用K均值聚类算法是数据挖掘技术在银行客户细分中一种重要方法,K均值算法存在对初始值敏感且容易陷入局部最优值的缺点,导致银户客户分类准确率低。为了提高银行客户细分的准确率,提出了一种基于改进的K均值聚类的银行客户细分方法。算法首先通过有效指数法动态调整初始聚类数K,减轻了聚类结果对初始聚类数K的依赖,通过自适应最佳密度半径来确定聚类中心,降低聚类中心对分类结果的影响,加快聚类速度,最后通过初始聚类数K和聚类中心对银行客户进行细分。在C++语言平台上,采用某市银业的客户分类数据对算法进行实验,结果表明,算法有效地克服了传统K均值算法易陷入局部最优值,提高了客户分类准确率,聚类结果更加合理,为银行决策者提高有效的参考,并带来更多的收益。Research on the classification of bank customers.K-means clustering algorithm in data mining technology,is important method in practical applications,but it is sensitive to the initial values and easy to get into the local optimal values.which results the shortcomings of low classification accuracy.In order to improve the classification accuracy of bank customers,a method based on improved k-means clustering is proposed.Firstly,the method adjusted the initial clustering dynamically through effective index method,reduced the number of k clustering results of initial clustering,then determined the clustering centers by optimal density radiuses,reduced the clustering centers to affect the results of classification,and quickened the speed of clustering.On C＋＋ language platform,using customer classification data to test the proposed method,experimental results show that the algorithm can effectively overcome the shortcoming that traditional k-means algorithm is easy to get into the local optimal values,and improve customer classification accuracy.The clustering results are more reasonable and advantageous to different groups of service strategy policymakers.

关 键 词：K均值算法 客户细分 聚类分析 银行 

分 类 号：TP311.52[自动化与计算机技术—计算机软件与理论]