Using Gradient Boosting Machines (GBM) Algorithm to Enhance Mobile Money System Security

Abstract

Joshua Mortey

This study proposes a machine learning framework to enhance mobile money security through fraud detection. Motivated by increasing threats like fraud and unauthorized access, it evaluates Gradient Boosting Machine (GBM) algorithms XGBoost, LightGBM, CatBoost, and AdaBoost using a public mobile transaction dataset. After preprocessing, training, and testing, model performance is assessed using accuracy, precision, recall, F1-score, and AUC. Results show GBM models outperform traditional methods, with XGBoost and CatBoost achieving about 99% detection accuracy and high precision. These models demonstrate strong potential for real-time fraud detection, enhancing financial security, inclusion, and user trust. The study also suggests further optimization for resource-limited environments and evolving cyber threats.

PDF