Implementation of Interpolation in Credit Card Fraud Detection

In today’s world, a critical threat to the Banking and Finance sector as well as its customers is the occurrence of fraud in the credit card transactions. Detecting this fraud is extremely arduous as it forms only a small percentage of the total number of transactions. In this paper, an algorithm which uses artificial neural network to detect fraudulent transactions amongst numerous genuine ones has been proposed. The number of fraudulent transactions is very few in comparison with the genuine transactions which introduces skewness in the data and makes the task of fraud detection difficult. In order to reduce the skewness of the dataset, under-sampling and over-sampling techniques have been used. The algorithm has been tested on three different datasets. The results of all the datasets have been compared. Confusion matrix and ROC plots have been compared. Further to improve the classification, generative adversarial networks (GANs) have been used to interpolate the fraudulent data without duplication on one of the datasets. These results show that there is an improvement in classification by using GANs.