Key points are not available for this paper at this time.
The banking industry's metamorphosis through digitalization has unquestionably revolutionized accessibility and convenience for customers worldwide. However, this paradigm shift has ushered in a new era of challenges, most notably in the realm of cybersecurity. Conventional rule-based fraud detection strategies have struggled to keep pace with the rapid evolution of cyber threats, prompting a surge of interest in more adaptive approaches like unsupervised learning. Furthermore, the COVID-19 pandemic has exacerbated the issue of bank fraud due to the widespread transition to online platforms and the proliferation of charitable funds, which present ripe opportunities for exploitation by cybercriminals. In response to these pressing concerns, this study delves into the realm of machine learning algorithms for the analysis and identification of fraudulent banking transactions. Notably, it contributes scientific novelty by developing models specifically tailored to this purpose and implementing innovative preprocessing techniques to enhance detection accuracy. Utilizing a diverse array of algorithms, including Random Forest, K-Nearest Neighbor (KNN), Naïve Bayes, Decision Trees, and Logistic Regression, the study showcases promising results. In particular, logistic regression and decision tree models exhibit impressive accuracy and Area Under the Curve (AUC) values of approximately 0.98, 0.97 and 0.95, 0.94, respectively. Given the pervasive nature of banking fraud in our digital society, the utilization of artificial intelligence algorithms for fraud detection stands as a critical and timely endeavor, promising enhanced security and trust in the financial ecosystem.
Johora et al. (Sat,) studied this question.