Analysis of machine learning and deep learning techniques for credit card fraud detection in class imbalanced datasets
Shyam Bahadur, Sudhanshu Kumar Jha · Computational Methods in Science and Technology · 2024
Credit card usage is increasing these days, and it plays a crucial role in the e-commerce business. Credit card frauds (CCFs) are now becoming more frequent due to the increasing use of credit cards. Financial institutions and card holders face monetary losses. The major challenge faced in the detection of credit card fraud is the imbalanced nature of the available datasets. The available datasets contain much less information about fraudulent transactions than legitimate transactions. Various machine learning-based approaches, such as Random Forest, KNN, SVM, ANN, and some hybrid methods, are available in literature. However, due to the low accuracy and high false-positive rates of machine learning techniques, deep learning algorithms along with different sampling methods need to be applied to reduce monetary losses. In this study, extensive analysis of both ML and DL methods has been done to judge the accuracy of various classifiers using various metrics on class-unbalanced datasets. This study also aims to suggest some efficient fraud detection algorithms that improve accuracy and reduce false positives.