Enhancing Credit Card Fraud Detection Through Adaptive Model Optimization
Chao Yan, Jinyin Wang, Yuelin Zou, Yijie Weng, Yang Zhao, Zhuoying Li · 2024
Credit card fraud remains a persistent threat in electronic transactions, driving ongoing advancements in fraud detection methodologies. We proposed an innovative method to fortify credit card fraud check systems through adaptive model optimization. We first start by addressing the inherent challenges of accurately defining fraud and managing imbalanced datasets, stressing the necessity for dynamic solutions to adapt to evolving fraud patterns. Our methodology leverages historical credit card transaction data to construct a flexible predictive model capable of discerning fraudulent activities while minimizing false classifications. By prioritizing the maximization of recall alongside controlling the false positive rate, our approach ensures robust fraud detection capabilities without sacrificing precision. Furthermore, we introduce novel techniques for feature analysis and preprocessing in the absence of comprehensive metadata, facilitating deeper insights into transaction characteristics. Through empirical validation on real-world datasets, we demonstrate the effectiveness of our adaptive model optimization framework in improving fraud detection accuracy and scalability. By implementing logistic regression, random forest, and XGBoost with the SMOTE method, our approach empowers businesses with proactive fraud prevention measures, safeguarding financial assets, and bolstering consumer trust in electronic payment systems.