Abstract
The study examines the effectiveness, challenges, and best machine learning algorithms for detecting e-commerce fraud. This study uses a systematic literature review to evaluate the effectiveness of machine learning-based e-commerce fraud detection, identify challenges, and identify the most effective techniques. The study examinedliterature extracted from the ScienceDirect, Emeralds, Wiley, and Springer databases, identifying 29 publications from recognized journals from 2012–2022, filtered using limitations and quality assessment criteria, and assessing paper eligibility. This study reveals that machine learning significantly enhances the accuracy of detecting e-commerce fraud. Yet, there are a number of issues that need to be resolved before machine learning can be utilized to detect e-commerce fraud. Poorer-quality data distribution is the biggest challenge in detecting e-commerce fraud. In order to determine the best machine learning strategy, the model's accuracy was also evaluated, and it was discovered that random forests performed the best in terms of accuracy. This study increases theoretical contributions as a continuation of previous research relevant to the concept of machine learning in detecting fraud in e-commerce. Then, based on the random forest's greater precision, it provides practical advice to e-commerce firms as a basis for decision-makers to find a suitable machine learning technique for fraud detection.