An Intelligent ABCBNLS-Based Context-Aware Credit Card Fraud Detection with User Spending Trajectory Analysis Using HFMM

Abstract

Robust Credit Card Fraud Detection (CCFD) is fundamental in the context of the digital Financial Sector (FS) for averting fraudulent transactions. Nevertheless, none of the existing CCFD studies concentrated on the Category Sequence Deviation (CSD) detection and User Spending Trajectory (UST) investigation, thereby overlooking the hidden fraudulent behaviours. Therefore, a robust context-aware CCFD with UST analysis and CSD detection utilizing Adaptive Bounded Cubic Bump Neutrosophic Logic System (ABCBNLS) and Hidden Faddeeva Markov Model (HFMM) is proposed in this paper. Primarily, the user registers in the financial application, followed by digital signature creation. Then, to initiate the transaction, the registered users log in to the application. Here, the user authentication is carried out via Digital Signature Verification (DSV). Afterward, the transaction attributes are extracted; further, the extracted transaction attributes are subjected to the trained proposed CCFD model. The CCFD dataset is gathered, followed by data pre-processing, contextual feature extraction, and correlation investigation. Thereafter, the context-aware pattern is identified. Next, by utilizing HFMM, the UST analysis and CSD detection are performed. In the meantime, to examine the behavioural risk, the proposed ABCBNLS is utilized based on the contextual features. Lastly, by using the Transfer Deep Long Adaptive Hyper-Tangent Rademacher Short Term Memory (TDLAHTRSTM), the credit card fraud is detected with an accuracy of 98.99748%. Hence, the normal transactions are completed and further stored on the Blockchain (BC) through the Proof-of-Stake (PoS) protocol. Overall, the hidden fraudulent behaviours were efficiently investigated by the proposed approach with a perplexity of 1.423.