Hybrid Methods for Credit Card Fraud Detection Using K-means Clustering with Hidden Markov Model and Multilayer Perceptron Algorithm

The use of credit cards is fast becoming the most efficient and stress-free way of purchasing goods and services; as it can be used both physically and online. Hence, it has become imperative that we find a solution to the problem of credit card information security and also a method to detect fraudulent credit card transactions. Over the years, a number of Data Mining techniques have been applied in the area of credit card fraud detection. The focus of this paper is to model a fraud detection system that would attempt to maximally detect credit card fraud by generating clusters and analyzing the clusters generated by the dataset for anomalies. The major objective of this study is to compare the performance of two hybrid approaches in terms of the detection accuracy. We employed hybrid methods using the K-means Clustering algorithm with Multilayer Perceptron (MLP) and the Hidden Markov Model (HMM) for this study. Our tests revealed that the detection accuracy of “MLP with K-means Clustering” is higher than the “HMM with K-means Clustering” for 80% percentage split but the reverse is the case when the “MLP with K-means Clustering” is compared with the “HMM with K-means Clustering” for 10 fold cross-validation but the accuracy is the same in the two hybrid methods for percentage split of 66%. More extensive testing with much larger datasets is however required to validate theses results.