Publication

Abstract : Credit card fraud is an essential problem in the economic industry; thus, its detection is solved with the help of the developed methods in order to minimize the overall loses and to improve the confidence of clients. Towards this end, this paper presents a novel framework that applies GNNs in combination with ensemble learning to accurately identify credit card fraud. GNNs are used for cardinality, representing various relationships between transactions, cardholders, and merchants where complicated types of forgery are identified and are beyond the capability of conventional approaches. Ensemble learning further improves the accuracy of the system in order to reduce high false positive rates by using multiple models to solve the problem of imbalanced classes that are characteristic of the COVID-19 data. The effectiveness of the proposed framework is thus compared with several state-of-the-art malicious account detection techniques and various highly relevant real credit card transaction datasets show the proposed framework attains better precision, recall, F1-score, scalability. Thus, this strategy is much more suitable for ongoing and unspecific problem addressing in an unobstructed and ambiguous environment, such as credit card fraud.