A Perceptron Based Neural Network Data Analytics Architecture for the Detection of Fraud in Credit Card Transactions in Financial Legacy Systems

Abstract

Credit card fraud, a significant and growing problem in commerce that costs the global economy billions of dollars each year, has kept up with technological advancements as criminals devise new and innovative methods to defraud account holders, merchants, and financial institutions. While traditional fraudulent methods involved card cloning, skimming, and counterfeiting during transactional processes, the rapid adoption and evolution of Internet technologies aimed at facilitating trade has given rise to new digitally initiated illegitimate transactions, with online credit card fraud beginning to outpace physical world transactions. According to the literature, the financial industry has used statistical methods and Artificial Intelligence (AI) to keep up with fraudulent card patterns, but there appears to be little effort to provide neural network architectures with proven results that can be adapted to financial legacy systems. The paper examines the feasibility and practicality of implementing a proof-of-concept Perceptron-based Artificial Neural Network (ANN) architecture that can be directly plugged into a legacy paradigm financial system platform that has been trained on specific fraudulent patterns. When using a credit checking subscription service, such a system could act as a backup.