节点文献
基于HMM模型的信用卡欺骗风险检测系统的仿真分析
【作者】 李洁；
【导师】 江雨燕；
【作者基本信息】 安徽工业大学 ， 管理科学与工程， 2010， 硕士
【摘要】 随着货币的电子化发展和电子商务的高速发展,信用卡在银行业务中所占的比例越来越大,同时,与信用卡相关的诈骗行为也越来越多,严重扰乱了正常的金融秩序,给银行和持卡人造成了很大的损失。本文仔细分析了信用卡(虚拟卡)交易的特点,发现隐马尔可夫模型能够有效的发现其内部隐藏的、潜在的异常交易模式。本文中构建的隐马尔可夫模型,将信用卡交易过程中不同的消费情况作为HMM模型中的随机过程,将交易金额的范围作为可观察到的输出概率矩阵,商品的类型作为模型中的状态,同时,利用客户的消费习惯来确定初始状态向量。使得HMM模型能够很好的应用到信用卡检测系统中。隐马尔科夫模型的建立是立足于持卡人的消费习惯的,训练阶段是脱机完成,检测阶段是在线进行的。并且在检测阶段不断修正,既能很好的检测出信用卡欺诈风险,又能适应持卡人消费习惯的改变。
【Abstract】 As the currency of electronic development and rapid development of e-commerce, credit cards in the banking business, a growing proportion of the same time, and the credit card-related fraud are also increasing, seriously disrupting the normal financial order, to the banks and caused great loss to the cardholder.This careful analysis of the credit card (virtual card) transaction characteristics, found that hidden Markov models can effectively discover its internal hidden potential unusual trading patterns.In this article build hidden Markov model, the process of credit card transactions on the spending of the different HMM models as a random process, the scope of the transaction amount, as can be observed in the output probability matrix, the type of goods as a model state, At the same time, the use of customer’s spending habits to determine the initial state vector. Makes the HMM model can be applied to credit card a good test system.Hidden Markov model is based on the cardholder’s spending habits, the training phase is completed offline detection stage is conducted on-line. And in testing phase of constant updating in a very good detection of both the risk of credit card fraud, but also to adapt to changes in cardholder spending habits.