Unsupervised Learning for Anomaly Detection in Financial Transactions

Abstract

The rapid growth of digital banking and online payment systems has resulted in a massive increase in financial transactions, making fraud and anomalous activities a serious concern. This project presents an Unsupervised Learning–based Anomaly Detection System for Financial Transactions, designed to identify unusual and suspicious transaction patterns without relying on labeled data. The system analyzes transaction behavior using machine learning techniques to learn normal patterns and detect deviations. A full stack web application is developed using Python (Flask) for backend processing and HTML, CSS, and JavaScript for frontend visualization. The platform provides interactive dashboards to display transaction insights and detected anomalies. The results demonstrate effective anomaly detection, real-time analysis, and improved interpretability, making the system suitable for financial monitoring and risk prevention.