Secure Credit Card Fraud Monitoring System
Anushri Saire1, Ashish Gawande2
1B. Tech Scholar, Department of Artificial Intelligence & Data Science, SBITM Betul (M.P)
2Assistant Professor, Department of Artificial Intelligence & Data Science, SBITM Betul (M.P)
Abstract - In today's digital age, fraud has increased significantly due to the widespread use of credit cards and the rapid growth of online trading. Credit card fraud detection is not only crucial for financial institutions but also important for ensuring customer trust and security. However, identifying fraudulent transactions in real-time remains a complex task due to the highly unbalanced nature of the data, where legitimate transactions are far outnumbered by fraudulent ones, and the need to reduce false positives to avoid blocking real users. The models investigated include logistic regression, decision tree, random forest, and extreme gradient boost algorithm (XG Boost). Suitable re sampling techniques, such as Overload Technology (SMOTE), are used to fix data issues in the data. Additionally, standard metrics are used to assess the recipient's standard performance, accuracy, recall, F1 score, and areas under the operating characteristic curve (ROC-AUC) to assess the effectiveness of each model. These models are cleverly improving to distinguish between fraudulent and legal transactions, while simultaneously maintaining a low false positive rate. The results suggest that integration of such an approach for machine learning in fraud detection systems can significantly improve performance and responsiveness. Ultimately, this paper presents the possibilities of advanced technology in machine learning to build more intelligent, faster, and more reliable fraud detection systems that can be adapted to the development of cyber threats in the financial domain: ensemble learning cyber security.
Key Words: Credit Card Fraud, Imbalanced Dataset, Logistic Regression, Decision Tree, Random Forest, XG Boost, SMOTE, ROC-AUC, F1-Score, Ensemble Learning.
