Online Transaction Anomaly Detection Model for Credit Card Usage Using Machine Learning Classifiers

Abstract

The methods for payment of an online transaction are credit or debit card, online bank transfer, e-wallets, mobile payments, etc. The process of electronic fund transfer is secure, and password protected, as claimed by the vendors. The security threats for all payment methods exist with an intention, like credit and debit card frauds. Monitoring fraudulent activities becomes more important as the number of online transactions for credit card usage grows over time. The detection of deviations from the large number of transactions in cases of frauds in credit card is desired using the classifiers of machine learning. We proposed to develop a Transaction Anomaly Detection (TAD) model for online transactions during credit card usage by customers using machine learning. The model is built to exploit and expose fraudulent transactions during online transactions at e-commerce sites. This work considers a data set from kaagle.com that has 28,4807 records of credit card transactions online with a class label. The proposed TAD model applies various machine learning algorithms to calculate the performance metrics and finds an efficient algorithm for detecting online transaction anomalies with good accuracy and recall. We observed that the XGB Classifier classifies the fraudulent transactions with an accuracy of 99.96% and a recall of 83%, which is the best suitable algorithm for this dataset.