RANDOM FOREST BASED FRAUD DETECTION METHOD FOR MULTI-PARTICIPANT E-COMMERCE TRANSACTIONS

Abstract

The primary goal of transactional security solutions has always been to detect
and prevent fraudulent transactions on e-commerce platforms. Due to the anonymity of online
transactions, it is difficult to identify attackers by just looking at past order data. Academics are
busy trying to come up with fraud prevention systems, but they haven't thought about how
consumers' behaviors are evolving. As a result, fraudulent behavior is not effectively detected.
An innovative approach to real-time user activity monitoring for fraud detection is presented by
this study, which combines process mining with algorithms grounded in machine learning. A
process model with user behavior detection is first developed for the business-to-consumer
online store. Secondly, we provide an anomaly-based approach to data mining that might be
applied to event logs. A classification model that employs SVM (support vector machine)
techniques to identify fraudulent activity is then fed the collected characteristics. The results of
the studies show that our technique successfully identifies dynamic fraudulent behavior on ecommerce
platforms.