Behavioral Pattern Analysis for Return Fraud Detection in High-Volume E-Commerce: A Multi-Signal Approach
DOI:
https://doi.org/10.63282/3050-9262.IJAIDSML-V7I1P148Keywords:
Return Fraud Detection, E-Commerce Fraud, Behavioral Anomaly Detection, Natural Language Processing, Multi-Signal Fusion, Risk Scoring, Cost-Sensitive Learning, Concept Drift, Imbalanced Classification, AI-Assisted Fraud DetectionAbstract
Return fraud costs U.S. retailers $103 billion annually, yet academic fraud detection literature remains almost entirely focused on payment transaction fraud. This paper presents a multi-signal behavioral analysis framework for return fraud detection that fuses three complementary signal streams: transaction-level features, behavioral history profiles, and natural language processing of customer-supplied return reason text. A late-fusion stacking architecture combines independently trained branch classifiers through a gradient-boosted meta-learner, enabling modular updates as fraud patterns evolve. Cost-sensitive learning, calibrated to the financial value of individual transactions, ensures that the system's optimization objective aligns with actual business impact rather than raw classification accuracy. A five-component evaluation methodology employs temporal validation and example-dependent cost metrics to establish realistic performance benchmarks that account for concept drift and class imbalance inherent to high-volume retail environments. The proposed system targets greater than 90% sensitivity for high-severity fraud categories while maintaining greater than 85% specificity to protect legitimate customer relationships. By integrating NLP analysis of return justification text a signal channel absent from all existing fraud detection literature despite a documented 76% consumer embellishment rate this framework addresses a structural gap in both academic research and commercial practice.
References
[1] Appriss Retail and Deloitte, "2024 Consumer Returns in the Retail Industry," Appriss Retail, Dec. 2024. https://apprissretail.com/resources/2024-consumer-returns-report/
[2] National Retail Federation, “NRF and Appriss Retail Report: $743 Billion in Merchandise Returned in 2023,” NRF, Dec. 26, 2023. https://nrf.com/media-center/press-releases/nrf-and-appriss-retail-report-743-billion-merchandise-returned-2023.
[3] National Retail Federation and Happy Returns, "2025 Retail Returns Landscape," NRF, 2025. https://nrf.com/research/2025-retail-returns-landscape.
[4] Optoro, “With Returns Fraud & Abuse on the Rise, and 69% of Shoppers Admitting to Wardrobing, with 64% Doing So At Least Once a Month,” Optoro, 2024. https://www.optoro.com/returns-news/optoro-launches-returns-unwrapped/.
[5] National Retail Federation, "NRF and Happy Returns Report: 2024 Retail Returns to Total $890 Billion
[6] ," NRF, Dec. 5, 2024. https://nrf.com/media-center/press-releases/nrf-and-happy-returns-report-2024-retail-returns-total-890-billion
[7] A. Mutemi and F. Bação, “E-Commerce Fraud Detection Based on Machine Learning Techniques: Systematic Literature Review,” Big Data Mining and Analytics, vol. 7, no. 2, pp. 419–444, Jun. 2024, doi: https://doi.org/10.26599/BDMA.2023.9020023.
[8] Y. Zhu et al., "Modeling Users' Behavior Sequences with Hierarchical Explainable Network for Cross-domain Fraud Detection," in Proc. The Web Conf. 2020 (WWW '20), Apr. 2020, pp. 928–938, doi: https://doi.org/10.1145/3366423.3380172.
[9] G. Zhang et al., “eFraudCom: An E-commerce Fraud Detection System via Competitive Graph Neural Networks,” ACM Transactions on Information Systems, vol. 40, no. 3, pp. 1–29, Jul. 2022, doi: https://doi.org/10.1145/3474379.
[10] A. Dal Pozzolo, G. Boracchi, O. Caelen, C. Alippi and G. Bontempi, “Credit Card Fraud Detection: A Realistic Modeling and a Novel Learning Strategy,” IEEE Transactions on Neural Networks and Learning Systems, vol. 29, no. 8, pp. 3784–3797, Aug. 2018, doi: https://doi.org/10.1109/tnnls.2017.2736643.
[11] Y. Dou, Z. Liu, L. Sun, Y. Deng, H. Peng, and P. S. Yu, “Enhancing Graph Neural Network-based Fraud Detectors against Camouflaged Fraudsters,” Proc. 29th ACM Int. Conf. Inf. Knowl. Manag. (CIKM '20), Oct. 2020, doi: https://doi.org/10.1145/3340531.3411903.
[12] Y. Liu et al., “Pick and Choose: A GNN-based Imbalanced Learning Approach for Fraud Detection,” Proceedings of the Web Conference 2021, Apr. 2021, doi: https://doi.org/10.1145/3442381.3449989.
[13] N. Tax et al., “Machine Learning for Fraud Detection in E-Commerce: A Research Agenda,” Deployable Machine Learning for Security Defense, vol. 1482, pp. 30–54, 2021, doi: https://doi.org/10.1007/978-3-030-87839-9_2.
[14] J. Yang, K. Chen, K. Ding, C. Na, and M. Wang, “Auto Insurance Fraud Detection with Multimodal Learning,” Data Intelligence, vol. 5, no. 2, pp. 388–412, Feb. 2023, doi: https://doi.org/10.1162/dint_a_00191.
[15] K. Taneja, J. Vashishtha, and S. Ratnoo, “Fraud-BERT: transformer based context aware online recruitment fraud detection,” Discover Computing, vol. 28, no. 1, Feb. 2025, doi: https://doi.org/10.1007/s10791-025-09502-8.
[16] V. R. Saddi, B. Gnanapa, S. Boddu, and J. Logeshwaran, “The Role of Natural Language Processing in Detecting Insurance Fraud,” 2023 4th International Conference on Communication, Computing and Industry 6.0 (C216), pp. 1–6, Dec. 2023, doi: https://doi.org/10.1109/c2i659362.2023.10430658.
[17] Z. Zhang et al., “Identifying E-Commerce Fraud Through User Behavior Data: Observations and Insights,” Data Science and Engineering, vol. 10, pp. 24–39, Jan. 2025, doi: https://doi.org/10.1007/s41019-024-00275-6.
[18] A. Correa Bahnsen, D. Aouada, A. Stojanovic, and B. Ottersten, “Feature engineering strategies for credit card fraud detection,” Expert Systems with Applications, vol. 51, pp. 134–142, Jun. 2016, doi: https://doi.org/10.1016/j.eswa.2015.12.030.
[19] A. C. Bahnsen, A. Stojanovic, D. Aouada and B. Ottersten, "Cost Sensitive Credit Card Fraud Detection Using Bayes Minimum Risk," 2013 12th International Conference on Machine Learning and Applications, Miami, FL, USA, 2013, pp. 333-338, doi: https://doi.org/10.1109/icmla.2013.68.
[20] J. Gama, I. Žliobaitė, A. Bifet, M. Pechenizkiy, and A. Bouchachia, “A survey on concept drift adaptation,” ACM Computing Surveys, vol. 46, no. 4, Article 44, pp. 1–37, Mar. 2014, doi: https://doi.org/10.1145/2523813.
[21] 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, Jun. 2002, doi: https://doi.org/10.1613/jair.953.
[22] A. Fernández, S. García, F. Herrera, and N. V. Chawla, “SMOTE for Learning from Imbalanced Data: Progress and Challenges, Marking the 15-year Anniversary,” Journal of Artificial Intelligence Research, vol. 61, pp. 863–905, Apr. 2018, doi: https://doi.org/10.1613/jair.1.11192.
[23] T. Saito and M. Rehmsmeier, “The Precision-Recall Plot Is More Informative than the ROC Plot When Evaluating Binary Classifiers on Imbalanced Datasets,” PLOS ONE, vol. 10, no. 3, p. e0118432, Mar. 2015, doi: https://doi.org/10.1371/journal.pone.0118432.
