| Abdullah Güney Işik, Şeyma Girgin, Hazal Su Biçakci Yeşilkaya Machine Learning-Based Credit Card Fraud Detection: Balanced and Imbalanced Cases Evaluation |
|---|
| Abstract. Credit card fraud has become an increasing threat with the development of the digital economy.brown In this study, different machine learning methods are examined by addressing the class imbalance problem in order to detect credit card frauds. The dataset is public and its name is a credit card transaction provided by Kaggle and represents real-world data. The training data is balanced by SMOTE (Synthetic Minority Over-sampling Technique) method, and the model performance is increased by extracting new features. Logistic Regression, Random Forest and XGBoost algorithms are applied and the performance of each is compared in terms of accuracy, precision, recall, F1 score and ROC/AUC values. In addition, the test data is evaluated both as imbalanced and balanced sample sets. The results show that the XGBoost model shows high performance especially in balanced test data. |
| Keywords: credit card fraud detection, online transactions, machine learning, SMOTE, classification |
 Download PDF |
| DOI: |
