A Machine Learning Framework for Credit Card Fraud Detection Using Random Forest and SMOTE

The rapid proliferation of online payment systems has made credit card fraud a critical financial security challenge. Detecting fraudulent transactions is inherently difficult due to severe class imbalance and continuously evolving fraud patterns. This paper presents a machine learning-based credit card fraud detection system integrated with a real-time web application. The system is built using a Python Flask backend and an HTML/JavaScript frontend. Uploaded transaction datasets are processed through missing value handling, outlier removal using the Interquartile Range (IQR) method, and class balancing via SMOTE. A Random Forest classifier is then dynamically trained on the processed data to classify transactions as fraudulent or legitimate. The application delivers rich visualizations including confusion matrix, precision-recall curve, fraud probability distribution, and transaction analysis graphs, along with a manual prediction module for real-time transaction testing. Experimental results demonstrate effective fraud detection with high accuracy, providing a practical, interpretable, and user-friendly fraud detection platform that combines machine learning and modern web technologies.