Hybrid Machine Learning-Based Multi-Stage Framework for Detection of Credit Card Anomalies and Fraud

Recently, tremendous growth in e-business has arisen in an increasing number of online transactions. Such widespread adaptation of e-payments has been going along with the increase in deceitful activities, which results in tremendous losses in the financial sector. This led to a novel research paradigm using statistical and auto-data-driven techniques to detect anomalies and fraud. Thus, traditional techniques fail to provide a secure medium for online transactions. Consequently, building a credit card fraud (CCF) detector is essential for secure online operations. Therefore, based on the abovementioned constraints, this paper presents a comprehensive study incorporating heterogeneous machine learning (ML) techniques for CCF detection. The proposed framework utilizes a multi-stage classification system that employs multiple classifiers, i.e., logistic regression, support vector machine (SVM) XGBoost, Random Forest, K-Nearest Neighbors (KNN), and Deep Neural Network (DNN). Furthermore, to accomplish the intensive class imbalance, the proposed technique uses a sampling technique with an internal features selection technique implemented based on voting among different methods. The key finding indicates that the proposed model surpasses the existing DNN simple voting, traditional stacking framework with a fraud recall value of 0.901, a legitimate recall value of 0.995, and a model cost value of 0.421.