基于时间序列孤立点检测的可疑外汇资金交易识别研究

发布时间：2018-01-28 11:56

本文关键词： 反洗钱可疑金融交易识别数据挖掘孤立点检测　出处：《统计与决策》2010年18期 　论文类型：期刊论文

【摘要】：尽管洗钱模式复杂多变,但洗钱行为在整个金融活动中只占有极少的比例,这给监测洗钱交易增加了难度。作为数据挖掘重要方法之一的孤立点分析是在大数据集中发现有趣小模式的有效方法。文章提出了一种适用于可疑外汇资金交易识别的孤立点检测方法,可以持续地从大量的日常交易中发现极少数的与正常交易显著不同的异常交易。从孤立点分析角度,提出了基于非频繁模式挖掘思想和概念漂移处理的混合属性空间上时间序列孤立点检测方法;从可疑金融交易识别的角度,提出了对每天持续动态产生的海量金融交易数据进行分析的一种新思路。
[Abstract]:Although the pattern of money laundering is complex and changeable, the behavior of money laundering occupies only a small proportion in the whole financial activities. This makes it more difficult to monitor money laundering transactions. As one of the important methods of data mining, outlier point analysis is an effective method for discovering interesting small patterns in big data concentration. This paper proposes a new method for suspicious foreign exchange fund transactions. The method of outlier detection. A small number of unusual transactions that are significantly different from normal transactions can be found continuously from a large number of daily transactions. Based on the idea of infrequent pattern mining and concept drift processing, a method for detecting outliers in time series in mixed attribute space is proposed. From the point of view of suspicious financial transaction recognition, a new way of analyzing the massive financial transaction data produced continuously and dynamically every day is put forward.
【作者单位】：西安交通大学经济与金融学院;
【分类号】：F830.92;F224
【正文快照】： 0引言关于“孤立点”,至今还没有一个正式的、被人们普遍接受的定义。中文存在孤立点、离群点、奇异点、异常点、偏离点等多种称谓。英文文献中存在Outlier、Anomaly、Unusualness、Novelty、Deviant、Noise、Exception等叫法。在本文中,统一称其为孤立点,英文Outlier。Hawkin

【参考文献】