Journal of Industrial Engineering Research in Production Systems

Journal of Industrial Engineering Research in Production Systems

A Multimodal Transformer for Retail Demand Forecasting Based on the Semantics of Product Descriptions and Purchase History

Document Type : Research Paper

Author
Fatemeh Zare Baghiabad
Department of Industrial Engineering, Faculty of Engineering, Ardakan University, P.O. Box 184, Ardakan, Iran
10.22084/ier.2026.32880.2238
Abstract
Accurate daily demand forecasting at the customer-product level remains a critical yet challenging task in retail supply chain management. This challenge is mainly driven by sparse transaction records, highly volatile consumer behavior, and the limited use of unstructured product information in forecasting models. To address these issues, this study proposes a novel multimodal semantic Transformer framework that integrates semantic product embeddings, extracted from product descriptions through natural language processing, with structured customer purchase history. In addition, the model incorporates multi-scale temporal features to capture seasonality, short-term demand fluctuations, and longer-term purchasing patterns more effectively. The proposed framework is evaluated on a real-world online retail dataset and compared with several benchmark models, including Long Short-Term Memory (LSTM) networks, Gradient Boosting Machines, and a unimodal Transformer architecture. The empirical results show that the proposed model consistently outperforms all baseline methods and achieves a 15.5% reduction in Mean Squared Error (MSE). The findings further reveal that semantic product information provides valuable predictive signals, particularly for low-velocity products where historical demand patterns alone are often insufficient. Moreover, temporal embeddings combined with dynamic attention conditioning improve the model’s ability to represent seasonal variations and evolving customer preferences over time. Overall, the results confirm that deep multimodal integration offers an effective, scalable, and interpretable solution for fine-grained retail demand forecasting and can support more customized inventory decisions and more resilient, customer-centric retail supply chains.
Keywords
Multimodal Transformer
Retail Demand Forecasting
Semantic Integration
Natural Language Processing (NLP)
Deep Learning
Inventory Management
Subjects
[1]    R. Tiwari, (2025). Harnessing AI and Predictive Analytics for Enhanced Demand Forecasting in Retail Supply Chains, Supply Chain & Retail Management, vol. 8, no. 1, 41–52.
[2]    L. Saarinen and P. Huttunen, (2025). Revisiting the value of data sharing in retail supply chain demand planning, International Journal of Operations & Production Management, vol. 45, no. 11, 1910–1936.
[3]    J. Oliveira and P. Ramos, (2024). Evaluating the effectiveness of time series transformers for demand forecasting in retail, Mathematics, vol. 12, no. 17, 2728.
[4]    M. Eskandari, M. Darabi, and H. Asghari, (2025). Identification and Assessment of Risks in the Sales Barriers to Reduce Food Supply Chain Crises, Journal of Supply Chain Management, vol. 27, no. 1, 18–29.
[5]    M. Sukel and M. Worring, (2024). Multimodal Temporal Fusion Transformers Are Good Product Demand Forecasters, IEEE Multimedia, vol. 31, no. 2, 48–60.
[6]    Y. Yan and N. Resnick, (2024). A high-performance turnkey system for customer lifetime value prediction in retail brands, Quantitative Marketing and Economics, vol. 22, no. 2, 169–192.
[7]    J. Rahikka and P. Mikkola, (2025). Modern time series methods for demand forecasting in retail.
[8]    A. R. Chowdhury, R. Paul, and F. Z. Rozony, (2025). A Systematic Review of Demand Forecasting Models for Retail E-Commerce Enhancing Accuracy in Inventory and Delivery Planning, International Journal of Scientific and Interdisciplinary Research, vol. 6, no. 1, 1–27.
[9]    T. Samal and A. Ghosh, (2025). Ensemble-based predictive analytics for demand forecasting in multi-channel retailing, Expert Systems with Applications, vol. 2025, 130212.
[10] Q. Li, (2023). Achieving Sales Forecasting with Higher Accuracy and Efficiency: A New Model Based on Modified Transformer, Journal of Theoretical and Applied Electronic Commerce Research, vol. 18, no. 4, 1990–2006.
[11] Y. Wang, (2025). Causal-Aware Multimodal Transformer for Supply Chain Demand Forecasting: Integrating Text, Time Series, and Satellite Imagery, IEEE Access, vol. 13, 176813–176829.
[12] W. Cai, Y. Song, and Z. Wei, (2021). Multimodal Data Guided Spatial Feature Fusion and Grouping Strategy for E-Commerce Commodity Demand Forecasting, Mobile Information Systems, vol. 2021, 5568208.
[13] X. Bi, G. Adomavicius, W. Li, and A. Qu, (2022). Improving Sales Forecasting Accuracy: A Tensor Factorization Approach with Demand Awareness, INFORMS Journal on Computing, vol. 34, no. 3, 1644–1660.
[14] R. M. A. et al., (2025). A Hybrid Temporal Convolutional Network and Transformer Model for Accurate and Scalable, IEEE Open Journal of the Computer Society, vol. 6, 380–391.
[15] S. Tripathi, R. Trigunait, and D. Chandra, (2025). A Behavioral and Environmental Framework for Sustainable Retail Forecasting and Decision-Making, Circular Economy and Sustainability, 1–29.
[16] S. K. Rai, (2025). Data-Driven Retail: The Engineering Behind Personalized Customer Experiences, Journal of Computer Science and Technology Studies, vol. 7, no. 10, 571–581.
[17] N. Tarighat et al., (2025). Domain Adaptation for Retail Demand Prediction, IEEE Access, vol. 13, 146267–146294.
[18] ف. زارع باقی‌آباد، (۱۴۰۵). پیش‌بینی پویای تقاضای خرده‌فروشی مبتنی بر رویکرد ترنسفورمر با تلفیق اثرات شبکه‌ای، تأخیرهای زمانی و معناشناسی محصول, مطالعات مدیریت صنعتی, دوره ۲۴, شماره ۸۱, ۱-۳۶.
[19] A. Dolgui, M. Pashkevich, (2008). Demand forecasting for multiple slow-moving items with short requests history and unequal demand variance. International Journal of Production Economics, vol. 112, no. 2, 885-894.
Journal of Industrial Engineering Research in Production Systems
Volume 13, Issue 27
September 2026
Pages 173-185