数据挖掘在反洗钱系统中的应用

发布时间：2018-04-24 19:31

本文选题：反洗钱系统 + 风险等级模型　；参考：《电子科技大学》2014年硕士论文

【摘要】：全球经济一体化进程的快速发展使得金融领域的洗钱行为日益猖獗,极大地影响了经济的正常运行。世界各国都在积极制定相关法律法规,并要求各个金融机构针对洗钱行为开发相应的反洗钱系统。本文在这种背景下,以实际开发的银行反洗钱系统为依据,研究了反洗钱的相关模式,并将反洗钱风险等级模型研究成果应用于反洗钱系统。本文的研究内容着重于数据挖掘技术在反洗钱系统中的应用这一课题,提出了一种洗钱风险等级模型。洗钱风险等级模型使用数据挖掘技术中的归纳分类技术,本文设计了一种基于决策树归纳和规则归纳的组合分类器,洗钱风险等级模型利用构建的组合分类器对银行客户信息属性进行洗钱风险等级类型预测分类,分别将客户洗钱风险等级分为高风险、中等风险和低风险三种等级。同时针对洗钱风险等级模型建立的组合分类器进行相应的树剪枝优化和Adaboost提升优化。本文对洗钱风险等级模型进行实验评估,使用客户信息验证集数据对洗钱风险等级模型进行测试,选用合理的评估度量标准比较各种基分类器以及组合分类器的性能,实验结果表明该组合分类器与单个分类器模型比较,具有较高的预测分类准确性。本文分析了银行现有反洗钱系统的结构与各个功能组件之间的关系,分析了SMBC-AML反洗钱系统的功能,目前反洗钱系统中存在的问题,介绍了按照国家反洗钱法的规定对大额、可疑、重点可疑等客户的识别,进而将洗钱风险等级模型以模块化的设计方式整合进当前的银行反洗钱系统中。实践表明该洗钱风险等级模型具有良好的应用前景。
[Abstract]:With the rapid development of global economic integration, money laundering in financial field is increasingly rampant, which greatly affects the normal operation of economy. Countries all over the world are actively making relevant laws and regulations, and all financial institutions are required to develop the corresponding anti-money laundering system. Under this background, based on the bank anti-money laundering system developed in practice, this paper studies the relevant models of anti-money laundering, and applies the research results of anti-money laundering risk level model to the anti-money laundering system. This paper focuses on the application of data mining technology in anti-money laundering system and proposes a risk level model for money laundering. In this paper, a combined classifier based on decision tree induction and rule induction is designed. The money laundering risk rating model uses the combined classifier to predict and classify the bank customer information attributes into three categories: high risk, medium risk and low risk. At the same time, the combined classifier based on money laundering risk grade model is optimized by tree pruning and Adaboost upgrading. This paper carries on the experimental evaluation to the money laundering risk grade model, uses the customer information verification set data to carry on the test to the money laundering risk grade model, selects the reasonable appraisal metric to compare the performance of various base classifier and the combination classifier. The experimental results show that the combined classifier is more accurate than the single classifier model. This paper analyzes the relationship between the structure of the existing anti-money laundering system of the bank and each functional component, analyzes the function of the SMBC-AML anti-money laundering system, the problems existing in the current anti-money laundering system, and introduces the large amount of money in accordance with the provisions of the National Anti-Money-Laundering Law. The identification of suspicious and key suspicious customers, and then the risk level model of money laundering is integrated into the current bank anti-money laundering system by modularized design. Practice shows that the model has a good application prospect.
【学位授予单位】：电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F832.2;TP311.13

【共引文献】