TY  - JOUR
T1  - Employing Support Vector Machines to Detect Credit Card Fraud for New Card Users
AU - , Rong-Chang Chen AU - , Chia-Jung Lin AU - , Li-Jung Lai AU - , u-Er Chien 
JO  - Asian Journal of Information Technology
VL  - 4
IS  - 2
SP  - 223
EP  - 228
PY  - 2005
DA  - 2001/08/19
SN  - 1682-3915
DO  - ajit.2005.223.228
UR  - https://makhillpublications.co/view-article.php?doi=ajit.2005.223.228
KW  - 
AB  - Credit card frauds cost cardholders and issuing banks hundreds of millions of dollars each year. It is, therefore, of great importance to use an effective method to prevent fraud. Employing a model based on the other cardholders` previous transaction data to detect whether a transaction is fraud or not has become a common solution to solve the fraud problem. However, for new users, this approach may not be appropriate because each user has different consumer behavior and thus it generally leads to low prediction accuracy. In contrast to the traditional approach, this study employs questionnaire-responded transaction (QRT) model to detect fraud for new users. First, an online questionnaire system is set up to gather the QRT data. Second, the QRT data are trained by using the support vector machines (SVMs) and the QRT models are produced. Finally, the QRT models are used to predict whether a new transaction is fraudulent or not. Experimental results from this study show that for new users, the QRT approach can detect the credit card fraud accurately.
ER  -