FRAUD SHIELD AI INTELLIGENT FINANCIAL FRAUD DETECTION AND PREVENTION SYSTEM USING MACHINE LEARNING

Abstract

The rapid growth of digital financial transactions has led to a surge in fraudulent activities, necessitating intelligent and adaptive detection systems. This paper presents Fraud Shield AI — a production-grade, real-time financial fraud detection and prevention system built using Python, FastAPI, scikit-learn, and an event-driven streaming architecture. The system ingests payment transaction events via a Kafka-based pipeline, applies a trained machine learning classifier to score each transaction for fraud probability, and exposes predictions through a low-latency REST API with a live web dashboard. A baseline Random Forest classifier is trained on the Kaggle Credit Card Fraud dataset and deployed as a FastAPI service containerised with Docker. MLflow is integrated for experiment tracking and model metadata management, while Prometheus and Grafana provide real-time monitoring of prediction throughput and service health. The system is evaluated on two dataset schemas and demonstrates high accuracy in distinguishing fraudulent from legitimate transactions. The proposed platform provides a reproducible, extensible baseline for real-time fraud detection research and illustrates end-to-end MLOps practices applicable to the financial services domain.