AI-Enhanced Security: The Future of Credit Card Fraud Detection

The study looks at the big issue of credit card theft, which has gotten worse because more people are shopping online, especially during the COVID-19 pandemic, when these thefts have increased a lot. The research uses methods like Logistic Regression, Decision Tree Classifier, and K-Nearest Neighbors. It analyzes data that includes changed numerical features from PCA analyses, as well as transaction time and amount. In fraud detection, the cross-validation score, ROC AUC score, and F1 score show how accurate these algorithms are. The cross-validation score for detecting fraud with a decision tree was 96.67% The ROC AUC score was 91.36%, and the F1 score was $\mathbf{9 0 \%}$. The K-Nearest Neighbors method worked really well for finding fraud. It had a ROC AUC score of $\mathbf{9 7}.63 \%$, a cross-validation score of $\mathbf{9 9.} 34 \%$, and an F1 score of 97%.