Title: OWDRN: optimised, weighted, and distributed recurrent neural network for product recommendation in e-commerce

Authors: Bharati Wukkadada

Addresses: Department of Datascience and Technology, K J Somaiya Institute of Management, Somaiya Vidyavihar University, Vidyavihar (E), Mumbai, Vidyanagar, Vidya Vihar East, Vidyavihar, Mumbai, 400077, Maharashtra

Abstract: Product recommendation suggests the intended and most useful appealing products based on customers' preferences and personalised experience. Based on these characteristics, several studies were conducted and ended with certain disadvantages as scalability, sparse data, cold start issues, computational complexities, and generalisability problems. These aforementioned problems are significantly addressed by proposing a model named optimised, weighted, and distributed recurrent neural network (OWDRN) for product recommendation. Further, the OWDRN model captures the sequential dependencies of user-item interactions using attention layers and embedding that provide better prediction accuracy. Additionally, the model improves performance by Menura honey optimisation (MHO), which reduced the local optima issues and achieved a better convergence rate specifically. Meanwhile, the distributed nature of the model allows the scale efficiency across multiple systems, ensuring robustness and accuracy in various conditions. Thus, the OWDRN model achieves high performance, with an accuracy of 96.84%, f1-score of 96.84%, precision of 97.21%, and recall of 96.48% under TP 90, offering highly personalised product recommendations.

Keywords: product recommendation system; distributed neural network; e-commerce; social networking; nature-inspired optimisation.

DOI: 10.1504/IJISCM.2025.154293

International Journal of Information Systems and Change Management, 2025 Vol.15 No.3, pp.290 - 320

Received: 04 Feb 2025
Accepted: 12 Aug 2025
Published online: 19 Jun 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article