Evolutionary Neuro-Fuzzy Network and Novel Hybrid Adaptive Crow Search-Modified Glowworm Swarm Optimization for Credit Card Fraud Detection

The growing usage of credit cards for online transactions is accompanied by a rise in fraud instances, fostering the need to develop advanced and interpretable models for fraud detection. To address this problem, a novel methodology is proposed using an Evolutionary Neuro-Fuzzy Network (ENFN) for optimal feature selection followed by a novel Hybrid comprising of Adaptive Crow Search Algorithm and Modified Glowworm Swarm Optimization Algorithm (hACSA-mGSO), seamlessly integrated with SVM classifier for robust classification. The ENFN employs adaptive mutation and fuzzy logic activations to optimize feature selection. The proposed hybrid metaheuristic algorithm (hACSA-mGSO) harnesses the global search ability of CSA and the local optimization capability of GSO for enhanced convergence. Adaptability is incorporated into the CSA through dynamic parameter updates along with stochastic modifications and random spiral movement is introduced in GSO to promote diversity among glowworms, mitigating premature convergence and facilitating efficient search space exploration. Through experimental evaluations using the credit card fraud detection dataset from the ULB Machine Learning Group, the proposed approach showcases its prowess in achieving high accuracy and efficient performance, thereby providing an efficient solution to the intricate challenge of credit card fraud detection.