Artificial Intelligence for Audit, Forensic Accounting, and Valuation

CHAPTER 5Machine Learning

CREDIT CARD BUSINESS is full of trepidations. Tired of fraud, many banks simply considered fraud as the necessary cost of doing business. Obviously, the losses from fraud are passed on to the customers, forcing banks to be less competitive in certain markets. Reducing fraud, therefore, is a direct driver of the bottom‐line value creation and of enabling banks to be more competitive. Less fraud means less cost, and that translates into being able to offer more competitive credit cards to customers. Many banks are now tackling the problem of fraud by applying machine learning. Visa is one of the companies that decided to conquer the challenge by developing a deep learning – based fraud detection system (Castellanos, 2019). However, while this was not the first time that a fraud detection system was used by a bank, this was one of the first attempts to use advanced AI for that purpose. Visa's older systems were based upon more basic machine‐learning and rules. Even though that system protected Visa from $25 billion of fraud attempts in the year ending in 2019, growing sophistication among the fraudsters required an even better system. The new system is a deep learning neural network. With higher sensitivity, the system can detect suspicious behavior with significantly higher precision.

The presence of fraud detection in credit card business is one application of deep learning in a firm. One can assume that based upon the success of that project, Visa ...