An Empirical Study of Methods, Metrics and Evaluation of Data Mining Techniques in Credit Card Fraudulence Detection

J. Karthika and A. Senthilselvi

Credit cards hold a major role in the current economy. It has become inevitable in many of the households, businesses and global purchases. Although this provides a large range of economic benefits, utmost care is to be taken when using it as these are subjected to greater fraud risks. Data Mining is recognized as one of the major technique for many disclosures of frauds especially on e-commerce. It also helps in identifying the risk factors of getting cheated. There are large volumes of data generated through the transactions made through a credit card and the bankers have the challenge of identifying the malicious and fraudulent transactions. The traditional methods fail in the process as it takes more time for analysing. Hence, Data mining techniques can be used for fraud detections. There are several methodologies for detecting the fraudulent usage of a credit card like Decision trees, Kmeans, neural networks, HMM and Genetic Algorithms. These methods increase the accuracy in the detection scenarios and reduce the time complexity. This paper investigates the various data mining techniques that are made used for the detection of credit card fraudulence with the metrics and evaluation comparison. The paper is aimed to give an insight on the available methods for CCFD.

Volume 12 | Issue 7

Pages: 351-362

DOI: 10.5373/JARDCS/V12I7/20202016