Machine Learning-Based Data Generative Techniques for Credit Card Fraud-Detection Systems

This study investigates the pressing issue of credit card fraud in the context of evolving e-commerce platforms and the necessity for improved fraud detection mechanisms. Since the advent of credit cards, the surge in their usage has led to a corresponding increase in fraud rates, highlighting the need to establish strong detection systems to prevent such activities. This research proposes a novel approach by integrating two distinct credit card datasets and a comparative evaluation of four machine learning imputation techniques to address missing values. By leveraging machine learning algorithms and imputation methods, we aim to enhance the accuracy and reliability of fraud detection. Our findings reveal significant improvements in model performance, with the accuracy of the integrated dataset reaching 100%, representing a 6.05% improvement over the original datasets; this improvement was confirmed to be statistically significant. Using the CBPM method, we selected the model that best balances accuracy and time efficiency. This result emphasizes the importance of effective data integration and imputation in combating financial fraud. It has direct practical implications for financial institutions, regulators, fraud analysts, and financial policymakers, who can use this approach to increase detection efficiency, reduce false positives, and optimize decision-making processes. Consequently, the method also helps protect consumers and enhances the overall resilience and credibility of financial markets.