Enhanced framework for credit card fraud detection using robust feature selection and a stacking ensemble model approach

Credit card fraud is an emerging global issue that causes substantial financial losses and undermines consumer trust in digital transactions. With the increase in online payment volumes, conventional fraud detection technologies are increasingly confronted by the complexity of fraudulent strategies that require intelligent and scalable alternatives. This study introduces an innovative machine learning-based fraud detection framework that incorporates sophisticated preprocessing methods like SMOTE-ENN for class imbalance mitigation, autoencoder for dimensionality reduction, and TOPSIS for optimal feature selection. A stacking ensemble model is developed with support vector machine (SVM), K-nearest neighbors (KNN), and extreme learning machine (ELM) to enhance forecast accuracy. The particle swarm optimization (PSO) algorithm is employed to optimize ELM parameters, enhancing generalization and model convergence. Extensive tests with standard datasets show outstanding results, achieving 99.95% accuracy, 99.93% precision, and 99.97% recall in detecting fraud. The outcomes highlight the model's proficiency in properly identifying fraudulent transactions while reducing false positives. The proposed method provides a viable alternative for secure and efficient credit card fraud detection in the contemporary digital economy, characterized by high accuracy and real-time scalability. • Propose a novel ML framework using advanced preprocessing and optimization to boost fraud detection accuracy. • Uses SMOTE-ENN, Autoencoder, and TOPSIS to balance data, reduce dimensions, and select optimal features. • Combines SVM, KNN, and PSO-optimized ELM to boost robustness, prevent overfitting, and speed up convergence.