International Journal of Computing and Digital Systems | 2026
Authors: Alahoum O.; Boudaa B.; Djafri L.; Abouaissa A.
DOI: 10.12785/ijcds/1571114787
Journal: International Journal of Computing and Digital Systems
Year: 2026
Publisher: University of Bahrain
Document Type: Article
Open Access: All Open Access; Gold Open Access
Cited by: 0
Recommendation systems (RS) are crucial for offering personalized suggestions based on user preferences. Recently, session-based recommendation systems (SBRS) have gained attention for their capability to model short-term user interactions. Graph neural networks (GNNs) have further improved SBRS by capturing complex relationships among items within sessions. In streaming environments, where user preferences change rapidly, traditional SBRS models are not adaptable. to cater for short-and long-term user demands. This study presents a GNN-based Streaming Session-Based Recommendation (SSBR) technique. To tackle the challenges posed by high-velocity data streams, it employs a reservoir-based active sampling strategy that efficiently updates the model while preserving key information. Furthermore, a Memory-Augmented Neural Network (MANN) is incorporated into the proposed model to enhance data-stream retention and adaptation. By combining MANNs with reservoir sampling, the model adapts its learning process, continually exploring changing user preferences while utilizing historical insights to deliver precise recommendations in real-time settings. Comprehensive tests on real datasets confirm that the proposed model achieves significant improvements over baseline approaches in terms of Mean Reciprocal Rank (MRR) and Recall across diverse settings. This evidence demonstrates its effectiveness in the SSBR setting. © 2026, University of Bahrain. All rights reserved.
Data Stream; Dynamic User Preferences; Memory-Augmented Neural Network; Streaming Session-Based Recommendation