一种聚类算法改进及其在税源分析中的应用  

Improved clustering algorithm and its application in revenue source analysis

作　　者：郑垂勇[1] 徐利[1,2] 淦文燕[3] 赵敏[1] 

机构地区：[1]河海大学商学院,江苏南京210098 [2]江苏省扬州市国税局,江苏扬州225000 [3]解放军理工大学指挥自动化学院,江苏南京210007

出　　处：《解放军理工大学学报（自然科学版）》2010年第2期202-206,共5页Journal of PLA University of Science and Technology(Natural Science Edition)

基　　金：江苏省哲学社会科学研究基金资助项目(09SJB790021)

摘　　要：面对大量的涉税数据,为揭示纳税个体间内在的分类特性,发现其中偏离常规的异常纳税行为等,有效提高税源监管力度,针对税源监管的应用需求,分析了数据挖掘中常用聚类算法的特性,提出了基于DENCLUE密度聚类的改进方法。该方法不仅可以发现数据分布的典型聚簇模式,而且可以揭示对税源监管更有价值的离群模式。应用实践验证了该方法的有效性。To expose the classified features of taxpayers as well as the abnormal ratepaying behaviors and to enhance the supervision, features of common clustering algorithm in data mining were analyzed according to the application requirements. An improved method based on DENCLUE density clustering was proposed. The new method could uncover not only the typical clustering mode of data distribution but the outlier mode as well. The experiment in the revenue source analysis shows the validity of this algorithm.

关 键 词：数据挖掘 聚类 税源分析 应用 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]