Cost sensitive modeling of credit card fraud using neural network strategy

摘要

Due to the rapid growth in e-business and electronic payment systems, Fraud is rising in banking transactions associated with credit cards. This paper intends to develop a credit card fraud detection (CCFD) model based on Artificial Neural Networks (ANN) and Meta Cost procedure to reduce risk reputation and risk of loss. ANN strategy have been used for credit card fraud prevention and detection. Because of the unbalanced nature of the data (Fraud and Non-Fraud cases), the detection of fraudulent transactions is difficult to achieve. To deal with the problem of imbalanced data, Meta Cost procedure is added. The proposed model, which is called Cost Sensitive Neural Network (CSNN), is based on misuse detection approach. Compared to the model based on Artificial Immune System (AIS), this model showed cost saving and increased detection rate. Data of this study is taken from real transactional data provided by a big Brazilian credit card issuer.