Review of Feature Selection Techniques for Credit Card Fraud Detection  
Author Huanjing Wang


Co-Author(s) Taghi M. Khoshgoftaar


Abstract Detecting credit card fraud is essential for financial institutions to prevent financial losses due to fraudulent transactions. Our research focuses on the Credit Card Fraud Detection Dataset, which is a widely used real-world transaction dataset, to identify the most crucial features that contribute to predicting fraudulent activities. Selecting the most relevant features helps reduce the amount of data that needs processing, saving time and computational resources. To overcome the common problems of high-dimensionality and class-imbalance in fraud detection datasets, we review a range of feature selection methods that are used either on their own or in combination with data sampling. Feature selection and data sampling have been successfully applied in credit card fraud detection, with tree-based feature selection methods such as Random Forest, Boruta, and Extra Tree classifier, as well as Synthetic Minority Oversampling Technique (SMOTE), being the most commonly used preprocessing techniques for this dataset.


Keywords Feature Selection, Credit Card Fraud, Class Imbalance, Data Sampling, Machine Learning
    Article #:  RQD28-396

Proceedings of 28th ISSAT International Conference on Reliability & Quality in Design
August 3-5, 2023