Research on Enhancing Credit Card Fraud Detection Based on a Comparative Study of Machine Learning Models and Imbalanced Data Strategies
Yiwei Li · Advances in Economics Management and Political Sciences · 2025
Global financial institutions are increasingly at risk from credit card fraud, which is expected to cause losses of over $43 billion by 2026. The efficacy of conventional rule-based fraud detection systems against complex and flexible fraud tactics has been steadily declining. In order to improve credit card fraud detection, this study investigates the use of artificial intelligence (AI), specifically machine learning models. This study applies an organised method that includes data pre-treatment, feature selection, and imbalance correction using SMOTE on a real-world transaction dataset. Five machine learning methods are compared: XGBoost, Random Forest, Decision Trees, Logistic Regression, and Gradient Boosting Decision Trees (GBDT). According to the findings, GBDT provides the best possible balance of recall, accuracy, and computing efficiency, which makes it ideal for real-time fraud detection systems in financial settings. Important factors including model interpretability, regulatory compliance, and the financial trade-offs between fraud prevention and false positives are also included in the paper. To create more resilient and flexible fraud detection systems, future prospects include combining behavioural analytics, reinforcement learning, and federated learning.