基于聚类算法的金融交易离群点识别  

作　　者：李欣月 张高煜[1] 彭兰舒 袁顺捷 常宇星 刘梦冬 

机构地区：[1]上海金融学院

出　　处：《电子技术（上海）》2016年第1期24-28,共5页Electronic Technology

基　　金：2014年上海大学生创新创业项目<云计算平台下的互联网反洗钱系统>(编号201411639039)基金支持

摘　　要：随着国家经济的逐渐发展,金融交易活动日渐频繁,交易产生的数据复杂而数量庞大,使得检测洗钱交易的进行变得困难,寻找能以较高准确度在海量数据中检测可疑金融交易算法成为关键。而洗钱金融交易行为相较于正常金融交易而言,发生的概率小且随机性强,处于与正常交易孤立的"离群"状态。针对这一特点,本文将在国内外现在有的关于金融可疑交易数据检测的研究基础上,提出建立在完整金融交易监控系统上的以聚类算法为基础的金融交易离群点检测算法,实现可疑金融行为的识别。With the gradual development of national economy and increasingly frequent financial transactions, data of financial transactions is growing complicated and volume of it becomes huge,thus letting detection of money-laundering transactions get difficult.Finding high accuracy algorithm for detecting suspicious financial transactions in huge amounts of data becomes critical.Compared with the normal financial transactions,money-laundering transactions have a small and random probability of occurrence, staying in isolated ＂ outlier＂ state.In order to achieve the recognition of suspicious financial activities,this paper will set up a algorithm for financial transaction outliers detection,which is based on Clustering algorithm and aimed at the characteristic of money-laundering transactions,combining the domestic and international existing researches on the detection of suspicious financial transaction data.

关 键 词：可疑金融交易识别 聚类算法 离群点 金融交易监控系统 

分 类 号：TN911.6[电子电信—通信与信息系统]