Hybrid Adaptive Sampling With Risk-based Credit Card Fraud Detection (has-rfd)

Abstract

Credit Card Fraud Has Become A Significant Challenge In The Digital Economy, Resulting In Substantial Financial Losses And Reduced Trust In Online Transactions. This Paper Presents A Hybrid Adaptive Sampling With Risk-Based Fraud Detection (HAS-RFD) Framework Integrated With Explainable Artificial Intelligence (XAI) Techniques To Improve Fraud Detection Performance. The Hybrid Sampling Method Combines SMOTE-based Oversampling With Clustering-based Adaptive Undersampling To Address Class Imbalance While Preserving Important Data Patterns. Machine Learning Models Such As Random Forest And Logistic Regression Are Trained On The Balanced Dataset To Classify Transactions As Fraudulent Or Legitimate. The System Incorporates A Risk-based Classification Mechanism That Categorizes Each Transaction Into Low, Medium, Or High Risk Levels. Explainability Techniques Provide Clear Insights Into Fraud Predictions, Enhancing Transparency And User Trust. The Proposed System Achieves Improved Detection Accuracy, Reduced False Positives, And Enhanced Interpretability, Making It Suitable For Real- World Financial Applications.

Keywords