基于SVDD的信用卡欺诈风险检测与管控研究

Research on Credit Card Risk Detection and Control Based on SVDD

摘要: 随着全球信用卡欺诈交易数量的与日俱增，由信用卡欺诈交易检测和预警不力所带来的银行及用户损失也日益增加。首先，本文通过判断欺诈检测准确性的TP和FP参数，运用概率估计确定算法训练集和测试集的容量。其次，以超球体为判断准则，运用拉格朗日函数训练得到最优超球面的半径和中心，建立信用卡欺诈消费的SVDD风险识别与评估模型。最后，提出信用卡欺诈的风险管控概率模型与反欺诈系统等处理措施。

Abstract: With the increase of credit card fraud transactions, the loss of banks and users caused by inade-quate detection and early warning untimely is also increased. Firstly, by the TP and FP parameters of the fraud detection accuracy, our paper determined the algorithm training set and test set through probability estimation. Then, using the hypersphere as a criterion, the Lagrange function was used to train the radius and center of the optimal hypersphere, and the SVDD risk identification and evaluation model for credit card fraud consumption was established. Finally, Credit card fraud risk control probabilistic models and anti-fraud systems were proposed.