Article Preview
TopIntroduction
Credit card fraud is a significant problem for banks worldwide, resulting in a massive loss of money. Technically, it is defined as the deliberate act of obtaining illegal benefits from users. Scammers today have taken advantage of the adoption of credit card systems as a primary online and offline mode of payment (Gamini, 2021). According to The Nilson Report, credit card fraud losses reached $21 billion globally in 2015 and are expected to hit $31 billion by 2020, of which at least 46% are victims from America between 2012 and 2016 (Taha & Malebary, 2020). In 2016, China also experienced a 3.8% increment in registered credit card fraud compared to the previous year. Also, according to the USFTC, identity fraud has increased by 21% in 2008, after remaining stable for the last few years (Rtayli & Enneya, 2020). Despite the challenges, credit card purchases have become commonplace in recent years, with the attribution of fraud to the increased use of internet credit cards in e-banking systems (Mittal & Tyagi, 2019). These threats have called for banks and card-related businesses to regularly step up their operations to identify credit card fraud identification to curb the menace (Wu, Xu, & Li, 2019; Kumari & Mishra, 2019). Though computational approaches are leveraged for the identification of fraud, the systems are data-dependent, and unfortunately, the data used for such tasks are significantly hampered by the data imbalance challenge (Darwish, 2020).
Several banks in the world today is concerned about protecting card payments and the general public's interest in making card payments (Gómez, Arévalo, Paredes, & Nin, 2018) as it is incredibly more convenient (Khatri, Arora, & Agrawal, 2020). In many advanced countries, credit cards are one of the most common payment methods for online transactions. It has made online purchases more accessible and useful as advanced technology such as the Internet of Things, mobile computing, and the Internet have evolved. However, it has also provided new opportunities for fraudsters and a unique challenge for implementers or innovators (Jain, Tiwari, Dubey, & Jain, 2019). To combat this issue, financial institutions use various fraud prevention models (Save, Tiwarekar, N., & Mahyavanshi, 2017), but these fraudsters are adaptable to the formulation of a different way to breach these protective models when given enough time. Fraudsters today can be a very imaginative, intelligent, and fast-moving group of people. Despite the best efforts of financial firms, law enforcement authorities, and the government, credit card fraud continues to grow.
Scientifically, credit card fraud detection systems are critical mechanism for preventing fraud incidents. The mechanism is usually divided into two categories: Anomaly detection and classifier-based detection. On the one hand, Anomaly detection is concerned with determining the distance between data points in space. They operate by filtering any incoming transaction which is inconsistent with the cardholder's profile while measuring the distance between them. On the other hand, the classifier-based detection approach employs machine learning techniques to train a classifier using supervised binary classification systems that have been adequately trained from pre-screened sample datasets (Zheng, Yan, Gou, & Wang, 2020).
The second approach is closely related to the data mining technique, and it is one of the most well-known tools for detecting credit fraud. Given adequate data, the method groups card transactions into two categories: legitimate (genuine) and fraudulent transactions. This decision is notoriously challenging due to field datasets' inherently imbalanced and distorted nature (Nadim, Sayem, Mutsuddy, & Chowdhury, 2019). The type of sampling method used, the variables chosen, and the detection technique(s) also used significantly has an impact on the efficiency of the credit card fraud detection system (Awoyemi, Adetunmbi, & Oluwadare, 2017; Kim et al., 2019; Rushin, Stancil, Sun, Adams, & Beling, 2017).