XAI Enabled Hybrid Model for Enhancing Financial Fraud Detection

At the ever-increasing deployment of digital financial services, security has become a concern between businesses and personal customers. Traditional fraud detection system mostly uses sophisticated machine learning models like Deep Neural Networks (DNN) and Recurrent Neural Networks (RNN) for suspicious transaction detection. This leads to compliance and trust issues. Moreover, centralization of the data is promoted by a number of fraud detection systems, which as a result becomes more susceptible to privacy and unauthorized access. To address the above restrictions, this research develops an Explainable AI (XAI)-focused fraud detection system which reserves both predictive performance and interpretable solutions. It uses Decision Trees, Random Forest and GBMs to show human-interpretable reasons about why a transaction is considered fraudulent. The effectiveness of XAI-led fraud detection models is evaluated using the Paysim1 dataset in this study. The bottom line is to design a fraud risk identification model that is both transparent, efficient, understandable and is able to protect the sensitive financial data at the same time.