Evaluation of Supervised Machine Learning Approaches for Credit Card Fraud Detection

Credit card fraud is one of the most common attacks affecting millions annually. Detecting fraud transactions using traditional software is not suitable anymore. The attackers are becoming more intelligent in their way of attacking. The investment of machine learning techniques in detecting suspicious activities is the trend nowadays. It helps stop fraud transactions without preventing benign transactions from being completed. This paper applies different supervised machine learning algorithms to a given dataset, such as random forest, gradient boosted trees, logistic regression, k-nearest neighbors, artificial neural network, and others. Consequently, the artificial neural network and the logistic regression algorithms show the highest performance according to the ROC-AUC performance measure. However, the k-nearest neighbor’s algorithm performs better according to the F1-score performance measure.

Перевод пока недоступен