Abstract:

Transaction security solutions have traditionally focused on identifying and stopping fraudulent transactions in e-commerce platforms. However, it is difficult to apprehend attackers using only the historical order information since e-commerce is hidden. Numerous studies attempt to create technologies that stop fraud, however they haven't taken into account consumers' changing behaviours from various angles. As a result, fraudulent activities are detected inefficiently. In order to do this, this study suggests a unique approach to fraud detection that combines process mining and machine learning models to track user behaviour in real time. We start by creating a process model for the B2C e-commerce platform that includes user behaviour detection. Second, a technique is described for examining anomalies in order to extract significant characteristics from event logs. A classification model based on Support Vector Machines (SVM) that can identify fraudulent activity is then fed the derived characteristics. Through the studies, we show how well our approach captures dynamic fraudulent behaviours in e-commerce platforms.

Description:

.