A Comparative Study of Machine Learning Classifiers for Credit Card Fraud Detection

Abstract

Now a day’s credit card transactions have been gaining popularity with the growth of e-commerce and shows tremendous opportunity for the future. Therefore, due to surge of credit card transaction, it is a crying need to secure it. Though the vendors and credit card providing authorities are showing dedication to secure the details of these transactions, researchers are searching new scopes or techniques to ensure absolute security which is the demand of time. To detect credit card fraud, along with other technologies, applications of machine learning and computational intelligence can be used and plays a vital role. For detecting credit card anomaly, this paper analyzes and compares some popular classifier algorithms. Moreover, this paper focuses on the performance of the classifiers. UCSD-FICO Data Mining Contest 2009 dataset were used to measure the performance of the classifiers. The final results of the experiment suggest that (1) meta and tree classifiers perform better than other types of classifiers, (2) though classification accuracy rate is high but fraud detection success rate is low. Finally, fraud detection rate should be taken into consideration to assess the performance of the classifiers in a credit card fraud detection system.