Beyond Traditional Methods: Evaluating Advanced Machine Learning Models for Superior Fraud Detection
Agampreet Singh, Kanwarpartap Singh Gill, Mukesh Kumar, Ruchira Rawat · 2024
The banking industry is increasingly concerned about credit card fraud due to its potential risks to both consumers and financial institutions. In recent years, machine learning techniques have proven effective in detecting fraudulent transactions. This study compares the performance of two wellknown machine learning methods such as XGBoost and Artificial Neural Networks (ANNs) in identifying credit card fraud. We assess the accuracy, precision, recall, and F1 score of these methods using a publicly available credit card transaction dataset. Additionally, the study examines the computational efficiency and scalability of ANNs and XGBoost to determine their suitability for real-time fraud detection systems. ANNs achieve the highest accuracy at 96.9%, surpassing all five methods evaluated, while XGBoost, with an accuracy of 92.7%, outperforms all other classifiers. These results provide valuable insights for financial institutions looking to implement or improve fraud detection systems, highlighting the strengths and limitations of each approach. This research contributes to ongoing efforts to leverage advanced machine learning techniques to combat credit card fraud and enhance financial security.