HOBA: A novel feature engineering methodology for credit card fraud detection with a deep learning architecture

Credit card transaction fraud costs billions of dollars to card issuers every year. A well-developed fraud detection system with a state-of-the-art fraud detection model is regarded as essential to reducing fraud losses. The main contribution of our work is the development of a fraud detection system that employs a deep learning architecture together with an advanced feature engineering process based on homogeneity-oriented behavior analysis (HOBA). Based on a real-life dataset from one of the largest commercial banks in China, we conduct a comparative study to assess the effectiveness of the proposed framework. The experimental results illustrate that our proposed methodology is an effective and feasible mechanism for credit card fraud detection. From a practical perspective, our proposed method can identify relatively more fraudulent transactions than the benchmark methods under an acceptable false positive rate. The managerial implication of our work is that credit card issuers can apply the proposed methodology to efficiently identify fraudulent transactions to protect customers' interests and reduce fraud losses and regulatory costs. (C) 2019 Elsevier Inc. All rights reserved.