<b>MACHINE LEARNING ENSEMBLES FOR ACCURATE DETECTION OF CREDIT CARD TRANSACTION FRAUD</b>

This paper focuses on the details of how to use machine learning ensemble approaches to detect this type of scam. The complexity of fraud schemes is on the rise, and remedies based on rules aren't working anymore. More advanced and flexible fraud detection tools are what we need now. To efficiently handle class mismatch in fraud datasets, enhance prediction accuracy, minimize variance, and use ensemble learning techniques like bagging, boosting, and stacking. These techniques employ numerous base models. The Kaggle credit card fraud dataset is one of the publicly accessible datasets used in this work to compare the performance of individual machine learning classifiers to that of several ensemble models, such as Random Forest, XGBoost, and Voting Classifiers. The objectives of the model evaluation are to reduce the occurrence of incorrect results and increase the rate of fraud detection. Area under the curve (AUC-ROC), F1-score, recall, accuracy, and precision were some of the evaluation criteria.