An evaluation of computational intelligence in credit card fraud detection

Abstract

Over the past few years, credit card transactions have been experiencing significantly fast development with the growth of e-commerce and shows tremendous promise of advancement in the future. Hence, due to explosion of credit card transaction, it is inevitable to secure transactions. To detect credit card fraud, computational intelligence has been widely used and plays a vital role. The selection of the preeminent classifier algorithm for a fraud detection system is a very extensive problem. This paper analyzes and compares various popular classifier algorithms that have been most commonly using in detecting credit card fraud. Moreover, it focuses on the measure used to assess the classification performance and rank those algorithms. Performance evaluation experiments were carried out on UCSD-FICO Data Mining Contest 2009 dataset. The experimental results suggest that (1) though achieved classification accuracy rate is 98.25% but fraud detection success rate is below 50%, (2) meta and tree classifiers perform better than other group of classifiers. Eventually, overall success rate (fraud detection rate) should be taken into consideration to evaluate the performance of the classifiers in a fraud detection system.