AN INTELLIGENT CREDIT CARD FRAUD DETECTION SYSTEM BASED ON GEOSPATIAL FEATURES AND ENSEMBLE MACHINE LEARNING

Abstract

As the global economy transitions rapidly towards digitization and cashless transactions, credit cards have become the primary medium of exchange. Consequently, the incidence of fraudulent activities has escalated exponentially, resulting in multi-billion dollar annual losses for financial institutions and consumers worldwide. Traditional static, batch-processing fraud detection systems struggle to adapt to the highly dynamic and contextdependent nature of modern transactional fraud. In this paper, we propose a comprehensive, real-time credit card fraud detection framework that integrates geospatial and occupational user behavior features with supervised machine learning models. We systematically evaluate six classification algorithms: Logistic Regression, K-Nearest Neighbors (KNN), Gaussian Naïve Bayes, Support Vector Machines (SVM), Decision Trees, and Random Forest Classifiers. The models are evaluated using actual transaction data containing geospatial coordinates and cardholder occupations. The empirical results demonstrate that the Random Forest and Decision Tree classifiers achieve a superior classification accuracy of up to 97.87%, with the Random Forest model achieving 100% recall on the test set. Furthermore, we deploy the predictive models within a real-time Django-based web application integrated with an SQLite database. This deployment enables instantaneous transaction evaluation, behavioral profiling, and administrative audit logging. The architecture successfully bridges the gap between theoretical machine learning models and practical, low-latency deployment in financial technologies.