A Multi-Perspective Fraud Detection Method for Multi-Participant Ecommerce Transaction

Abstract

It integrates Support Vector Machines (SVM) for fraud classification, Petri nets for anomaly detection, and process mining for conformance checking, ensuring a scalable and robust detection system.

Introduction

Overview

With the rapid growth of e-commerce, digital transactions are increasingly vulnerable to sophisticated fraud, resulting in global financial losses. Current systems struggle to analyze transaction processes from multiple perspectives, limiting their effectiveness. To address this, the study proposes a hybrid fraud detection approach combining machine learning and process mining for real-time and historical analysis of user behavior and transaction patterns.

Objectives

1. Identify and classify suspicious transactions based on behavioral anomalies.

2. Improve accuracy and reduce false positives in fraud detection systems.

Related Works

• [1] G. Mesnita (2019): Evaluated decision trees, neural networks for fraud detection.

• [2] Lutao Zheng et al. (2018): Used logical graph modeling for better transaction analysis.

• [3] Riyanarto Sarno et al. (2015): Integrated association rules with process mining to reduce false discoveries.

Proposed Method

A hybrid approach incorporating:

1. Process Mining: Uses conformance checking to detect anomalies in transaction flow.

2. Petri Nets: Capture and analyze user behavior in transaction sequences.

System Architecture

Key stages:

• Data Collection: Gathers transactional records and activity logs.

• Data Cleaning: Removes inconsistencies and prepares data for modeling.

• Algorithm Application: Applies machine learning (e.g., decision trees, neural networks) to detect fraud patterns.

• Model Building: Trains predictive models to detect fraudulent behavior.

• Prediction: Flags suspicious transactions in real-time or batches.

Workflow

• Users log in, view fraud predictions, analyze transaction legitimacy, and access datasets.

• Admin features allow viewing remote users.

• Ends with logout.

Sequence Models

• UML sequence diagrams represent interaction flows among system components during fraud detection processes.

UML Use Case Diagram

• Illustrates user interactions with the system.

• Shows extension relationships between use cases using UML standards.

Results and Discussion

• Utilized a Kaggle dataset with floating-point attributes and fraud labels.

• Model performance was evaluated based on classification accuracy.

• Findings stress the importance of classifier selection and suggest further enhancement through feature engineering or trying alternative algorithms.

Conclusion

The study introduces an effective multi-perspective fraud detection system that leverages the strengths of machine learning and process mining to improve precision, adaptiveness, and reliability in identifying fraudulent e-commerce transactions.

References

[1] R. A. Kuscu, Y. Cicekcisoy, and U. Bozoklu, Electronic Payment Systems in Electronic Commerce. Turkey: IGI Global, 2020, pp. 114–139 [2] P. Rao et al., “The e-commerce supply chain and environmental sustainability: An empirical investigation on the online retail sector.” Cogent. Bus. Manag. 1938377, 2021.

Copyright

Copyright © 2025 Mrs. Asha Jyothi Panchala , K Joshna, K Harika, M Guru Charan , Uday Kiran. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.