This paper discusses the performance of machine learning algorithms in detecting fraudulent credit card transactions. Machine learning is effective for anomaly detection. Naturally, it should also be effective for fraudulent credit card transactions. The focus here is on which algorithm is more suitable for this problem. In other words, this research aims to find machine learning techniques with high fraud detection performance. We tried seven major machine learning algorithms: Logistic Regression, Naive Bayes, Decision Tree, Random Forest, K-Nearest Neighbors, Gradient Boosting, and Support Vector Machine. In addition, since the proportion of fraudulent transactions is generally small, we combined three data sampling methods to adjust the amount of data in the fraud class: Synthetic Minority Over-sampling Technique (SMOTE), Random Over-sampling, and Random Under-sampling. The Borda method was used to evaluate the overall performance results from multiple perspectives. The experimental results revealed that the decision tree-based algorithms performed better overall. Random Forest and Decision Tree performed the best, followed by Gradient Boosting. Additionally, it was found that creating a training dataset with a more minor variance in the fraud amounts leads to better performance. Regarding sampling methods, over-sampling-based methods, SMOTE and Random Over, were superior. Furthermore, the Random Forest–SMOTE and Decision Tree–Random Over combinations were highly accurate for lower and higher fraudulent transaction amounts.