The XGBoost Approach Tuned by TLB Metaheuristics for Fraud Detection

Abstract

The recent pandemic had a major impact on online transactions. With this trend, credit card fraud increased. For the solution to this problem the authors explore existing solutions and propose an optimized solution. The solution is based on an extreme gradient boosting algorithm (XGBoost) and a teaching-learning-based-optimization algorithm. The dataset optimizes the hyperparameters of the XGBoost which is utilized as the main driver for the solution. The evaluation was performed among other similar techniques that have solved this problem successfully in the past. Standard performance metrics were applied which are accuracy, recall, precision, Matthews correlation coefficient, and area under the curve. The result of this research presents a dominant solution that was proposed and successfully outperformed all other compared solutions overall.