An Augmentation of Credit Card Fraud Detection using Random Undersampling

Abstract

The two most prominent necessities which triggered the advent and growth of Digital Transactions are ease of doing payments and security of the transactions. However, even after a lot of research into the field, financial cyber crime is still marred with lots of digital frauds and corruption; credit card fraud is one of them. A plethora of patents and research papers have tried to solve this issue, but the secure transaction is still a distant dream. As per a survey, $24.26 Billion was lost worldwide due to payment card fraud only in 2018. This study aims to augment the performance of credit card fraud detection using an undersampling based data analysis technique. The study used an ensemble method to detect credit card fraud. The data used for simulation was highly imbalanced. Hence, a random undersampling technique was applied to datasets to make it a balanced dataset. The validation of performance augmentation was done based on the predefined performance measure metrics such as accuracy, precision, recall, f-score, geometric-mean, and the area under curve score with the receiver operating curve. The main focus was to check the ensemble method’s performance on imbalanced data and improve it by providing it a balanced data. The study results showed that the augmentation of the detection technique of credit card fraud was improved with a random under-sampling method on a credit card transaction imbalanced dataset. The study performed a comparative analysis of the augmentation of detection technique of credit card fraud models before and after incorporating random under-sampling techniques on credit card fraud imbalanced datasets.