The AdaBoost Approach Tuned by Firefly Metaheuristics for Fraud Detection

Abstract

The use of powerful classifiers is broad and the problem of fraud detection tends to benefit from similar solutions as well. The problem in the digital age cannot be disregarded as the number of cases is worrisome. The use of machine learning has been beneficial to many real-world problems, as the classification ability of such solutions is high. Furthermore, these solutions are not without shortcomings, and possibilities of hybrid methods are explored for the reasons of further enhancements. Therefore, in the research proposed in this manuscript, the adaptive boosting algorithm is optimized by the firefly metaheuristics and validated against the imbalanced credit card fraud detection dataset. Moreover, the synthetic minority over-sampling technique is applied for addressing the class imbalance. According to experimental findings, the proposed method shows substantially better performance than other state-of-the-art machine learning models for tackling the same problem in terms of standard classification metrics.