A MULTI-PRESPECTIVE FRAUD DETECTION METHOD FOR MULTIPARTICIPANT ECOOMERCE TRANSACTIONS

Abstract

The primary goal of transaction security systems in e-commerce platforms has always been to identify and stop fraudulent transactions. But since online shopping is so covert, it's not possible to catch criminals using only past orders as evidence. A lot of studies have tried to come up with fraudprevention technology, but they haven't taken consumers' ever-changing behaviours into account. As a result, fraudulent behaviours are not effectively detected. In order to accomplish this goal, this article presents a new approach to detecting fraud by following users' actions in real-time using models from machine learning and process mining. To begin, we integrate user behaviour detection into a process model that pertains to the business-to-consumer e-commerce platform. Secondly, we provide an anomaly-based approach to feature extraction from event logs. After that, we use a classification model based on Support Vector Machines (SVMs) to identify fraudulent behaviours using the characteristics that were retrieved. Our technique successfully captures dynamic fraudulent behaviours in ecommerce systems, as shown in the studies.