AI-Driven Framework for Enhanced Fraud Detection in Financial Transactions

Department of Computer Applications Manipal University Jaipur Jaipur, India
Department of Computer Applications Manipal University Jaipur Jaipur, India
Department of Computer Applications Manipal University Jaipur Jaipur, India
Department of Computer Applications Manipal University Jaipur Jaipur, India
Department of Computer Applications Manipal University Jaipur Jaipur, India

This email address is being protected from spambots. You need JavaScript enabled to view it.
This email address is being protected from spambots. You need JavaScript enabled to view it.
This email address is being protected from spambots. You need JavaScript enabled to view it.
This email address is being protected from spambots. You need JavaScript enabled to view it.
This email address is being protected from spambots. You need JavaScript enabled to view it.

Abstract

Digital finance is becoming more widespread and with this, the probability of credit-card fraud is also growing, and this represents a grave aspect of the banking industry as well as its user. This paper suggests a practical approach to detecting fraud, which is supported by artificial intelligence (AI), combines machine-learning (ML) with deep-learning (DL)-based systems, and provides solid real-time protection. The suggested system integrates three major models, i.e. autoencoders, Random Forest, and XGBoost, to assess a range of attributes in transactional nature, such as monetary value, date, geographical coordinates, merchant facts, and regular user behaviour. Facing the ongoing challenge of the issue of class imbalance in the fraud data, approaches like Synthetic Minority Over-sampling Technique (SMOTE) as the means of oversampling and a one-hot encoding as the variable classification tool are utilized to increase the reliability of training the model. Instead of simply defining transactions as fraudulent or legitimate, our system gives each instance a score probability hence making it easy to intervene actively as a stakeholder. As a web-based dashboard, desired results are displayed in visual form with Receiver Operating Characteristics (ROC) curves and heatmaps, and model explanations explained using SHAP and LIME models, thus increasing transparency. Based on benchmark datasets like IEEE-CIS and the Credit Card Fraud repository on Kaggle, the system delivers the remarkable performance metrics in terms of accuracy, precision and recall. It is an application that is cloud-native, therefore, capable of reacting suitably to demand to maintain security cover at all times. All in all, the scalability of technologies and explainable AI facilitate fraud detection to me as slicker, more transparent, and more aligned with the demands of the contemporary financial institutions and their customers, which is a big leap towards safer virtual finance.