Synergistic Effects of Transactional Patterns and Demographic Profiles on Financial Fraud Detection Accuracy

Abstract

The rise of digital banking ecosystems and the Unified Payments Interface (UPI) has increased transactional efficiency, but it has also created new vectors for financial fraud. Current fraud detection methods mostly make use of either transaction behaviors or demographic information independent of each other, restricting their ability to consider the dependency between them. The study proposes a synergic fraud detection framework for better evaluation of financial fraud risk management and support of managerial decisions by using both transactional and demographic information. The fraud detection framework incorporates the behavioral analysis, statistical learning, clustering based segmentation and anomaly scoring techniques to derive Transactional Risk (T) score and Demographic Risk (D) score and then utilizes synergic interaction model (T × D) to incorporate their combination effect. The experimental results show the model outperforms existing transactional and demographic approaches in detecting the frauds and achieves a considerable improvement of 95% in accuracy transactional-only approaches equals to 87.5% and demographic-only approaches equals to 82%. Also, precision improves to 93.5% and the recall level becomes 95%. In addition, the false positive rate switches to 48%, the fraud detection rate achieves 48% and the effectiveness of risk segmentation attains 80% a reduction of 42.5% in cost of investigation versus 17% in transaction-based models and 11% in demographic models. However, the major contribution of this research paper is to prove that the fraud risk does not arise because of independent behavior and demographic variables but due to the interaction between these two aspects. The developed framework adds interpretability, increases the efficiency of risk segmentation process and helps in making instant decisions for managers in the BFSI setting.