基于嵌入法–卷积神经网络的信用卡欺诈检测研究

Application of Embedding Method-Convolutional Neural Network in Credit Card Fraud Detection

摘要: 随着电子银行的发展和在线支付方式的完善，信用卡的使用量和发行量在逐步增加，信用卡诈骗等欺诈行为与日俱增，造成极大的经济损失。对信用卡欺诈行为精准识别、快速检测已经成为当下研究的重点。为解决单一算法泛化能力不强，精度不高等问题，本文结合大数据背景，应用深度学习相关算法，研究信用卡客户的交易行为，构建嵌入法–卷积神经网络的信用卡欺诈检测数学模型，与单一算法模型相比，改善了过拟合，提高了准确度，降低了误分类概率，对金融机构信用卡欺诈检测具有较好的适用性和可行性。

Abstract: With the development of e-banking and the improvement of online payment methods, the use and circulation of credit cards are gradually increasing, and fraud such as credit card fraud is increasing day by day, resulting in great economic losses. Accurate identification and rapid detection of credit card fraud have become the focus of current research. Single algorithm to solve the generalization ability is not strong, the accuracy is not high. This article combines with background of big data, application of deep learning algorithms, studies the trading behavior of credit card customers, builds imbedding method-convolution mathematical model of the neural network credit card fraud detection, compares with the single algorithm model, improves the fitting and the accuracy, and reduces the probability of misclassification; it has good applicability and feasibility for credit card fraud detection in financial institutions.