Detection Banking Fraud Using Reinforcement Learning

Credit card fraud is a global issue that causes significant harm to both individuals and organizations. Each year, the losses from fraud in the banking sector continue to grow, highlighting the need for new approaches to detect and prevent it. In this context, tools that not only identify fraudulent transactions based on known patterns but also detect new forms of fraudulent behavior are particularly important. Therefore, considerable attention is being given to the application of machine learning methods for security tasks, including the detection of fraudulent activities.One such method is the use of reinforcement learning. The main goal is to understand how reinforcement learning can play a crucial role in detecting credit card fraud as user behavior patterns change. The proposed model employs Deep Q-Network (DQN) to detect fraud in real-time. A publicly available Kaggle dataset was used for training and testing the model. As a result, a web application and model were developed, demonstrating high effectiveness with an accuracy of 0.9992 and a precision of 0.8289.