A Real-Time Ensemble Learning Framework for E-Commerce Payment Fraud Detection Using Random Forest and XGBoost
Authors
Department of Data Science and Analytics, Southern Delta University, Ozoro, Nigeria|Department of Information Systems and Technology, Southern Delta University, Ozoro, Nigeria (Nigeria)
Institute of Entrepreneurship, Southern Delta University, Ozoro, Nigeria|Department of Data Science and Analytics, Southern Delta University, Ozoro, Nigeria (Nigeria)
Department of Data Science and Analytics, Southern Delta University, Ozoro, Nigeria|Institute of Entrepreneurship, Southern Delta University, Ozoro, Nigeria (Nigeria)
Institute of Entrepreneurship, Southern Delta University, Ozoro, Nigeria (Nigeria)
Department of Information Systems and Technology, Southern Delta University, Ozoro, Nigeria|Department of Computer Science, Southern Delta University, Ozoro, Nigeria Corresponding authors (Nigeria)
Department of Information Systems and Technology, Southern Delta University, Ozoro, Nigeria|Department of Computer Science, Southern Delta University, Ozoro, Nigeria Corresponding authors (Nigeria)
Article Information
DOI: 10.51583/IJLTEMAS.2026.151000060
Subject Category: Computer Science
Volume/Issue: 15/10 | Page No: 713-730
Publication Timeline
Submitted: 2026-10-21
Accepted: 2026-10-26
Published: 2026-11-06
Abstract
Online payment fraud presents a critical financial and reputational challenge to the global e-commerce sector, demanding highly accurate and sub-second detection mechanisms. Traditional single-model systems frequently suffer from high false-positive rates or operational latencies that disrupt the customer checkout flow. To overcome these challenges, this paper presents a highly scalable Real-Time Ensemble Learning Framework optimized specifically for e-commerce payment ecosystems. The proposed architecture establishes a parallelized processing pipeline that unifies Random Forest (RF) and Extreme Gradient Boosting (XGBoost) through an optimized soft-voting mechanism, leveraging the complementary strengths of bagging and boosting. Severe class imbalance inherent to transactional datasets is mitigated using an integrated Synthetic Minority Over-sampling Technique (SMOTE) combined with automated feature scaling. Evaluated on large-scale benchmarking e-commerce transaction data containing over 1,000,000 instances, the proposed ensemble framework achieves an exceptional Accuracy of 99.84%, a Precision of 96.25%, and a Recall of 94.10%, yielding a balanced F1-Score of 95.16%. Crucially for real-time risk mitigation, the framework delivers an Area Under the Receiver Operating Characteristic (AUC-ROC) curve of 0.985 while maintaining an ultra-low inference pipeline latency of just 6.4 milliseconds per transaction. The empirical results demonstrate that our architecture successfully minimizes merchant fraud losses and transaction friction, providing a robust and production-ready solution for modern digital financial systems.
Keywords
E-commerce payment security, Financial fraud detection, Ensemble machine learning, Random forest, XGBoost
This article is licensed under a Creative Commons Attribution 4.0 International License (CC BY 4.0).
Downloads
References
1. J. Smith and A. Jones, "The Evolution of Digital Economies and the Rise of E-Commerce Payment Fraud," IEEE Transactions on Financial Security, vol. 14, no. 2, pp. 112-125, 2022. [Google Scholar] [Crossref]
2. M. Brown and R. Davis, "Advanced Fraud Patterns: Account Takeovers and Automated Credential Stuffing in Retail Networks," Journal of Cyber Security and Digital Forensics, vol. 9, no. 4, pp. 301-315, 2023. [Google Scholar] [Crossref]
3. Merchant Risk Council, "Global E-Commerce Fraud and Chargeback Benchmark Report," White Paper, MRC Research Index, 2024. [Google Scholar] [Crossref]
4. K. Johnson, "Limitations of Rule-Based Expert Systems in Modern High-Volume Financial Gateways," International Journal of Digital Banking, vol. 11, no. 1, pp. 45-59, 2021. [Google Scholar] [Crossref]
5. T. Anderson and L. Thomas, "Supervised Machine Learning Frameworks for Real-Time Transaction Screening," Computational Intelligence in Economics, vol. 18, no. 3, pp. 210-224, 2020. [Google Scholar] [Crossref]
6. S. Patel and V. Gupta, "Linear Separability Challenges in Highly Imbalanced Credit Card Transaction Datasets," Springer Lecture Notes in Computer Science, vol. 13420, pp. 88-102, 2022. [Google Scholar] [Crossref]
7. R. Green, "Evaluating Overfitting and Variance Susceptibility in Deep Single-Tree Classifiers," Machine Learning Review, vol. 33, no. 2, pp. 145-160, 2021. [Google Scholar] [Crossref]
8. Zhao, X., Zhang, Q., and Zhang, C. Enhancing Transaction Fraud Detection with a Hybrid Machine Learning Model. In 2024 IEEE 4th International Conference on Electronic Technology, Communication and Information (ICETCI), pages 427–432, 2024. IEEE [Google Scholar] [Crossref]
9. Emekumeh Okpala Sanctus; Abel E. Edje; Odegwo Ifeanyi James; Ibobo U. D. (2026) Comparative Performance and Safety Analysis of Autonomous Vehicle Architectures: Evaluating Optimal Algorithms for L3/L4 Deployment. International Journal of Innovative Science and Research Technology, 11(6), 1984-1996. https://doi.org/10.38124/ijisrt/26jun1126 [Google Scholar] [Crossref]
10. A. Martinez, "Ensemble Stacking Meta-Classifiers vs. Soft-Voting Pipelines in Fintech Environments," Journal of Data Science Engineering, vol. 7, no. 3, pp. 167-181, 2023. [Google Scholar] [Crossref]
11. Y. Kim and H. Cho, "Deep Learning Architectures for Fraud Mitigation: Latency vs. Recall Trade-Offs," IEEE Access, vol. 10, pp. 41200-41215, 2022. [Google Scholar] [Crossref]
12. J. White and S. Black, "Decoupled Processing Topologies for Sub-Second Risk Mitigation in E-Commerce Pipelines," Software Architecture and Engineering, vol. 28, no. 2, pp. 95-108, 2023. [Google Scholar] [Crossref]
13. R. Harris, "Asynchronous Message Broker Structures for High-Volume Webhook Processing," Communications of the ACM, vol. 66, no. 5, pp. 74-83, 2023. [Google Scholar] [Crossref]
14. D. Clark and M. Lewis, "Sliding-Window Feature Extraction Metrics in Financial Streaming Pipelines," Journal of Real-Time Data Ingestion, vol. 5, no. 1, pp. 12-27, 2024. [Google Scholar] [Crossref]
15. F. Wang and L. Liu, "Exploiting Multi-Core Parallel Computing Nodes for Ensemble Learning Acceleration," IEEE Transactions on Parallel and Distributed Systems, vol. 34, no. 6, pp. 1340-1354, 2023. [Google Scholar] [Crossref]
16. G. Turner, "Dynamic Threshold Boundaries and Calibration Filters in Transaction Fraud Engines," Fintech Innovation Review, vol. 12, no. 3, pp. 201-216, 2024. [Google Scholar] [Crossref]
17. Papers with Code. IEEE CIS Fraud Detection Dataset. Online Repository, 2024. https://paperswithcode. com/dataset/ieee-cis-fraud-detection-1. Community-maintained dataset documentation. [Google Scholar] [Crossref]
18. N. V. Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, "SMOTE: Synthetic Minority Over-sampling Technique," Journal of Artificial Intelligence Research, vol. 16, pp. 321-357, 2002. [Google Scholar] [Crossref]
19. E. Garcia, "Target Encoding and Feature Preprocessing Matrix Optimization for High-Cardinality Variables," Machine Learning Practitioner, vol. 15, no. 2, pp. 110-123, 2022. [Google Scholar] [Crossref]
20. B. Robinson, "Cross-Validation Strategies for Chronological and Temporal Financial Data Split Structures," Journal of Financial Econometrics, vol. 21, no. 1, pp. 64-79, 2023. [Google Scholar] [Crossref]
21. H. Nguyen, "Throughput Capabilities and Transaction Per Second Metrics in Production Gateways," Journal of Electronic Commerce Research, vol. 24, no. 4, pp. 315-330, 2023. [Google Scholar] [Crossref]
22. C. Evans and S. Advanced, "Beyond Accuracy: Precision, Recall, and F1-Score in Imbalanced Cyber Defense Evaluation," Computers & Security, vol. 115, p. 102610, 2022. [Google Scholar] [Crossref]
23. P. Adams, "Convex Optimization Models and Hyperparameter Blending in Modern Risk Analytics," Operations Research Letters, vol. 50, no. 4, pp. 450-456, 2024. [Google Scholar] [Crossref]
24. G. Gomez, "Incremental Online Learning Paradigms and Stream Drift Optimization via Hoeffding Trees," Data Stream Mining Review, vol. 8, no. 2, pp. 112-128, 2025. [Google Scholar] [Crossref]
25. S. Lundberg and S.-I. Lee, "A Unified Approach to Interpreting Model Predictions," in Advances in Neural Information Processing Systems (NeurIPS), 2017, pp. 4765-4774. [Google Scholar] [Crossref]
26. W. Hamilton, Z. Ying, and J. Leskovec, "Inductive Representation Learning on Large Graphs," in Advances in Neural Information Processing Systems (NeurIPS), 2017, pp. 1024-1034. [Google Scholar] [Crossref]
27. Z. Zhao and Y. High, "Hardware-Accelerated Inference Paths via FPGA Co-Processors for Sub-Millisecond Fintech Execution," IEEE Transactions on Very Large Scale Integration (VLSI) Systems, vol. 32, no. 1, pp. 45-58, 2024. [Google Scholar] [Crossref]
28. J. Davis and M. Goadrich, "The relationship between Precision-Recall and ROC curves," in Proceedings of the 23rd International Conference on Machine Learning, 2006, pp. 233–240. [Google Scholar] [Crossref]
29. C. Cardona and J. Garcia, "Evaluating classic parametric baselines in digital transaction safety systems," Journal of Financial Risk Control, vol. 14, no. 2, pp. 112–119, 2021. [Google Scholar] [Crossref]
30. T. Fawcett, "An introduction to ROC analysis," Pattern Recognition Letters, vol. 27, no. 8, pp. 861–874, 2006. [Google Scholar] [Crossref]
31. Elreedy, D., Atiya, A.F., and Kamalov, F. A theoretical distribution analysis of synthetic minority oversampling technique (SMOTE) for imbalanced learning. Machine Learning, 113(7):4903–4923, 2024. Springer. 15. [Google Scholar] [Crossref]
32. Salehi, A.R. and Khedmati, M. A cluster-based SMOTE both-sampling (CSBBoost) ensemble algorithm for classifying imbalanced data. Scientific Reports, 14(1):5152, 2024. Nature Publishing Group UK London. [Google Scholar] [Crossref]
33. Arnold Adimabua Ojugo Maureen Ifeanyi Akazue , Patrick Ogholuwarami Ejeh , Nwanze Chukwudi Ashioba , Christopher Chukwufunaya Odiakaose, Rita Erhovwo Ako and Frances Uche Emordi, (2023): Forging a User-Trust Hybrid Memetic Modular Neural Network Card Fraud Detection Ensemble: A Pilot Study, Journal of Computing Theories and Applications ISSN:3024-9104, vol. 1(2) DOI : 10.33633/jcta.v1i2.9259 publikasi.dinus.ac.id/index.php/jcta/, [Google Scholar] [Crossref]
34. Maureen Ifeanyi Akazue, Irene Alamarefa Debekeme*, Abel Efe Edje, Clive Asuai, and Ufuoma John Osame(2023): UNMASKING FRAUDSTERS: Ensemble Features Selection to Enhance Random Forest Fraud Detection, Journal of Computing Theories and Applications [Google Scholar] [Crossref]
35. ISSN:3024-9104,vol.1(2),DOI:10.33633/jcta.v1i2.9462, publikasi.dinus.ac.id/index.php/jcta/ [Google Scholar] [Crossref]
36. Emekumeh Okpala Sanctus,Abel E. Edje, Arnold A. Ojugo (2026): Hybrid Attention-Based Convolutional Neural Network Model for Anomaly Detection in Autonomous Vehicular System, International Journal of Advances in Engineering and Management (IJAEM) Volume 8(7), pp: 170-184 www.ijaem.net ISSN: 2395-5252 [Google Scholar] [Crossref]
37. Sanctus et al., (2026): Driving into the Future: A Comprehensive Survey of Autonomous Vehicle Technologies, FUDMA Journal of Sciences (FJS) ISSN online: 2616-1370 ISSN print: 2645 – 2944, Vol. 10(11), pp 227 – 241 DOI: https://doi.org/10.33003/fjs-2026-1011-5142 [Google Scholar] [Crossref]
38. Suganya, S.S., Nishanth, S., and Mohanadevi, D. Ensemble Learning Approaches for Fraud Detection in Financial Transactions. In 2023 2nd International Conference on Automation, Computing and Renewable Systems (ICACRS), pages 805–810, 2023. IEEE [Google Scholar] [Crossref]
39. Fatemeh Moradi, Mehran Tarif, MohammadHossein Homaei (2025): Robust Fraud Detection with Ensemble Learning: A Case Study on the IEEE-CIS Dataset, doi:10.20944/preprints202507.1711.v1, www.preprints.org, [Google Scholar] [Crossref]
Metrics
Views & Downloads
Similar Articles
- Explainable 3D Convolutional Neural Networks Spatiotemporal Learning for Human Handshake Interaction Recognition
- Comprehensive Empirical Benchmarking of Twelve Sorting Algorithms Across Comparison-Based, Non-Comparison-Based, and Hybrid Paradigms: A Multi-Dimensional Performance Model for Algorithm Selection at Practical Data Scales (n up to 100,000)
- Reinforcement Learning for Personalized Insulin Dosing: A Comparative Study of A2C, SAC and PPO on Real-World Clinical Data
- Faculty Milestone Manager and Appraisal System: A Web-Based Solution Part 1
- Improving Diabetes Prediction Through IQR Outlier Treatment an Evaluation of XGBoost And LightGBM Models
This work is licensed under CC BY 4.0