A Dual Approach for Credit Card Fraud Detection using Neural Network and Data Mining Techniques

The world is being more and more digitalized with each passing day, making information security a huge concern. Through efficient fraud detection using a robust approach, the banking industry will be able to avert fraudulent transactions and save millions of dollars every year. Each money transaction is crucial and thus, fraudulent transactions need to be detected at any cost. In this paper, we build models to detect fraudulent credit card transactions using five classifiers to find out the best fit classifier for the situation. We use two different techniques to tackle the inherent problem of data imbalance. The first technique uses data resampling technique to increase the number of samples in the minority class, whereas, the second one uses a cost-based approach where the error function incorporates weights of each class. Through the weights, more emphasis can be given to the samples of fraudulent transactions than the normal samples.