Home Journals:
Author Services About Submit Article


Academic Publishing

E-ISSN 3041-4849
European Journal of Computer Sciences and Informatics
« Previous Article
Next Article »

Research Article
Online Published: 12 Nov 2026
 


FraudGuard: An Intelligent Adaptive System for Real-Time E-Commerce Fraud Detection with SHAP-Based Explainability and Cost-Efficient Decision Making

Marwan Mosleh, Amani Yousef Owda, Majdi Owda.


Abstract
Aim/Background: Fraud is a major challenge in e-commerce that causes huge financial losses every year. Fraud detection is challenging because the patterns continue changing faster than static rules can be updated, and the traditional method fails to detect complex methods of fraudulent transactions.

Methods: In this paper, we proposed FraudGuard, an explainable hybrid system that combines rule-based with machine learning and SHAP-based explainability to detect fraud in e-commerce transactions. The rule is prioritized based on cost optimizations to improve efficiency. We evaluated the performance by testing five machine learning algorithms: logistic regression, random forest, balanced random forest, gradient boosting classifier, and XGBoost on 23,634 e-commerce transactions (5.17% fraud rate, 2025). We integrated SHAP (SHapley Additive exPlanations) to explain the machine learning predictions.

Results: Using cross-validated decision thresholds, the results showed that the Balanced Random Forest achieved optimal standalone performance with F1 = 0.322 (95% CI [0.278, 0.365]) with 0.463 recall and 0.246 precision. FraudGuard achieves 85.7% recall with 69.4 ms average latency, 26.8% faster than the ML-only approach, while resolving 36.2% of transactions through the rule layer alone and maintaining comprehensive SHAP-based explainability. We further integrate SHAP to automatically create new rules from machine learning results. In a 200-transaction test simulation only generated one new rule and reduced the ML call rate from 67.5% to 62.5%, but it did not show F1 improvement at this scale.

Conclusion: Precision remains as limitation and still low 10.8%, and ~89% of flagged transactions are false positives.

Key words: E-Commerce, Fraud Detection, Artificial Intelligence (AI), Hybrid System, Rule-Based Reasoning, Machine Learning, Decision Support


 
ARTICLE TOOLS
Abstract
PDF Fulltext
How to cite this articleHow to cite this article
Citation Tools
Related Records
 Articles by Marwan Mosleh
Articles by Amani Yousef Owda
Articles by Majdi Owda
on Google
on Google Scholar


How to Cite this Article
Pubmed Style

Mosleh M, Owda AY, Owda M. FraudGuard: An Intelligent Adaptive System for Real-Time E-Commerce Fraud Detection with SHAP-Based Explainability and Cost-Efficient Decision Making. Eu J Comp Sci Informatics. 2026; 3(4): 207-220.

doi:10.5455/EJCSI.20260531090727



Web Style

Mosleh M, Owda AY, Owda M. FraudGuard: An Intelligent Adaptive System for Real-Time E-Commerce Fraud Detection with SHAP-Based Explainability and Cost-Efficient Decision Making. https://www.wisdomgale.com/ejcsi/?mno=322770 [Access: November 14, 2026].

doi:10.5455/EJCSI.20260531090727



AMA (American Medical Association) Style

Mosleh M, Owda AY, Owda M. FraudGuard: An Intelligent Adaptive System for Real-Time E-Commerce Fraud Detection with SHAP-Based Explainability and Cost-Efficient Decision Making. Eu J Comp Sci Informatics. 2026; 3(4): 207-220.


doi:10.5455/EJCSI.20260531090727



Vancouver/ICMJE Style

Mosleh M, Owda AY, Owda M. FraudGuard: An Intelligent Adaptive System for Real-Time E-Commerce Fraud Detection with SHAP-Based Explainability and Cost-Efficient Decision Making. Eu J Comp Sci Informatics. (2026), [cited November 14, 2026]; 3(4): 207-220.


doi:10.5455/EJCSI.20260531090727



Harvard Style

Mosleh, M., Owda, . A. Y. & Owda, . M. (2026) FraudGuard: An Intelligent Adaptive System for Real-Time E-Commerce Fraud Detection with SHAP-Based Explainability and Cost-Efficient Decision Making. Eu J Comp Sci Informatics, 3 (4), 207-220.


doi:10.5455/EJCSI.20260531090727



Turabian Style

Mosleh, Marwan, Amani Yousef Owda, and Majdi Owda. 2026. FraudGuard: An Intelligent Adaptive System for Real-Time E-Commerce Fraud Detection with SHAP-Based Explainability and Cost-Efficient Decision Making. European Journal of Computer Sciences and Informatics, 3 (4), 207-220.


doi:10.5455/EJCSI.20260531090727



Chicago Style

Mosleh, Marwan, Amani Yousef Owda, and Majdi Owda. "FraudGuard: An Intelligent Adaptive System for Real-Time E-Commerce Fraud Detection with SHAP-Based Explainability and Cost-Efficient Decision Making." European Journal of Computer Sciences and Informatics 3 (2026), 207-220.


doi:10.5455/EJCSI.20260531090727



MLA (The Modern Language Association) Style

Mosleh, Marwan, Amani Yousef Owda, and Majdi Owda. "FraudGuard: An Intelligent Adaptive System for Real-Time E-Commerce Fraud Detection with SHAP-Based Explainability and Cost-Efficient Decision Making." European Journal of Computer Sciences and Informatics 3.4 (2026), 207-220. Print.


doi:10.5455/EJCSI.20260531090727



APA (American Psychological Association) Style

Mosleh, M., Owda, . A. Y. & Owda, . M. (2026) FraudGuard: An Intelligent Adaptive System for Real-Time E-Commerce Fraud Detection with SHAP-Based Explainability and Cost-Efficient Decision Making. European Journal of Computer Sciences and Informatics, 3 (4), 207-220.


doi:10.5455/EJCSI.20260531090727