XGBoost based solutions for detecting fraudulent credit card transactions

Due to the emigration of businesses to the internet, credit cards have become a widely used tool for customers to pay for purchases both online and offline. However, fraudsters try to attack these payment tools using various techniques, and credit card fraud transactions have become dangerous. Credit card payment security is becoming an important research topic, and different ways are proposed to overcome this problem. In this article, we will propose a solution that combines the strengths of three methods to secure credit card fraud transactions. The Differential Evolution (DE) algorithm is used to select the best hyperparameters of the XGBoost algorithm (eXtreme Gradient Boosting). Synthetic Minority Oversampling (SMOTE) combined with Edited Nearest Neighbour (ENN) to handle the imbalanced data issues. Finally, we will classify fraud transactions using the optimized XGBoost algorithm. The proposed solution is evaluated using the European data set in terms of Accuracy, Recall, F1score, and AUC. As a result, the proposed solution generally achieved the best score in terms of all the measurements used in this study. In terms of accuracy, we achieved 99.94%, similarly, we reached the best score in terms of the Precision score, which is 80.68%, in addition, we got 86.02% of recall score, 83.27 of F-measure, and 99.21 of AUC score. Thus, we emphasize the superiority of our proposed method over the competitive machine learning model utilized in this work.