Fundamentals and Methods of Machine and Deep Learning

12Application of Machine Learning Algorithms With Balancing Techniques for Credit Card Fraud Detection: A Comparative Analysis

Shiksha

School of Computer Science and Mathematics, Liverpool John Moores University, Liverpool, United Kingdom

Abstract

This paper investigates the performance of the four supervised machine learning algorithms: logistic regression, support vector machine, decision tree, and random forest on the highly imbalanced ULB machine learning credit card fraudulent transaction dataset and a comparative analysis is performed. The major point that distinguishes this study from others is that this work is not only focused on the four supervised machine learning algorithms, rather than the permutation and combination of these methods with different balancing techniques are studied and analyzed for several evaluation metrics to obtain a better performance. In addition of the sampling techniques, one more method is used to balance the dataset by taking in account the balanced class weight at the modeling time. It is important to mention that the random forest with balanced class weight has shown the lowest false positive transactions with a value of only 3. The comparative results demonstrate that the random forest with SMOTE oversampling technique has output the best results in terms of all the selected metrics with accuracy of 99.92%, recall value of 81.08%, precision of 76.43%, F1-score of 78.69%, MCC of 0.79, AUC of 0.96, and AUPRC of 0.81.

Keywords: Supervised ...