Design and Empirical Analysis of Data-Driven Machine Learning Models for Credit Card Fraud Detection

As credit card usage becomes increasingly widespread, fraudulent activities pose a growing threat to financial security. This study aims to develop efficient credit card fraud detection models using machine learning algorithms. Three algorithms-logistic regression, Support Vector Machine, and decision trees-were employed to analyze a dataset comprising transaction records from 10,000 users. The results demonstrate that all models performed exceptionally well in terms of precision, exceeding 98%. Notably, logistic regression exhibited the best overall accuracy and recall, while decision trees showed advantages in reducing false positives. These findings provide valuable insights for financial institutions, enabling them to select appropriate fraud detection methods based on specific requirements. Despite limitations in data scale and the proportion of fraudulent samples, this research offers significant contributions to the field of credit card fraud detection, laying the groundwork for constructing a more secure financial environment. Future research directions may focus on expanding the dataset, exploring deep learning algorithms, and developing real-time fraud detection systems.