Credit Card Fraud Detection Using Modern Machine Learning And Deep Learning Approaches

The rapid shift toward digital and card-based payment systems has intensified the challenge of detecting fraudulent financial transactions. Static, rule-oriented detection techniques are often unable to recognize newly emerging fraud strategies. This work introduces a data-driven credit card fraud detection framework that leverages machine learning and deep learning algorithms to improve detection accuracy. Transaction records are prepared through feature normalization and class rebalancing using the Synthetic Minority Oversampling Technique (SMOTE). Several supervised classifiers, including Logistic Regression, Random Forest, and XGBoost, are trained and comparatively analyzed. Experimental evaluation highlights the superior performance of ensemble-based models in identifying fraudulent activity. The selected model is deployed through a lightweight Flask web application to enable real-time transaction assessment, making the solution practical for real-world financial systems.