Year: 2025
Type: article
Abstract: In recent decades, the rise of E-commerce has made credit cards a popular choice for transactions due to their convenience. However, this has also led to an increase in credit card fraud, posing a significant financial threat. To tackle this issue, this study investigates the use of machine learning models for credit card fraud detection. The research explores three models: Gradient Boosting, Decision Trees, and K-Nearest Neighbors (KNN), assessing their performance using metrics like Area under the Precision-Recall Curve (AUPRC), precision, recall, and F1-score. The study employs the Synthetic Minority Oversampling Technique (SMOTE) algorithm to address the dataset's class imbalance, creating a balanced training dataset. Among the models, Gradient Boosting achieves the highest AUPRC of 0.81, indicating its effectiveness in general scenarios. However, its precision in detecting fraudulent transactions remains a challenge. The Decision Tree model provides a balanced performance with an AUPRC of 0.59, while KNN stands out with a higher precision for fraudulent transactions, achieving an AUPRC of 0.77. These findings highlight the need for continuous improvement in fraud detection techniques, suggesting that hybrid models or ensemble approaches could enhance detection capabilities by leveraging the strengths of different models.
DOI: 10.54097/5bxeae69
Author: Manho Yeung
Institution: University of California, Santa Barbara
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4407682331