Article: Dynamic modelling of brand-user relationships via graph neural networks for enhanced marketing optimisation Journal: International Journal of Information and Communication Technology (IJICT) 2025 Vol.26 No.39 pp.74 - 91 Abstract: To address the limitations of existing methods in modelling dynamic and heterogeneous brand-user relationships, this study proposes a dynamic heterogeneous graph attention network (DHGAT). This framework integrates three core innovations: 1) a time-decay-based edge weighting mechanism that quantifies temporal dynamics of user-brand interactions; 2) a cross-relation attention layer that distinguishes semantic differences among diverse behaviours (e.g., purchases vs. complaints) through relation-specific gating; 3) a reinforcement learning decision engine optimising marketing actions via Q-learning. Validated on a real-world e-commerce dataset (32,000 users, 142M interactions), DHGAT achieves an AUC of 0.892 in relationship prediction (5.7%-16.8% higher than baselines) and boosts marketing ROI by 41% in online A/B tests. The framework enables end-to-end optimisation of marketing strategies while balancing short-term conversions and long-term user value, offering a novel paradigm for data-driven marketing decision systems. Inderscience Publishers - linking academia, business and industry through research

Title: Dynamic modelling of brand-user relationships via graph neural networks for enhanced marketing optimisation

Authors: Aoxue Yao

Addresses: School of Digital Communication, Guangzhou Huashang College, Guangzhou 511300, China

Abstract: To address the limitations of existing methods in modelling dynamic and heterogeneous brand-user relationships, this study proposes a dynamic heterogeneous graph attention network (DHGAT). This framework integrates three core innovations: 1) a time-decay-based edge weighting mechanism that quantifies temporal dynamics of user-brand interactions; 2) a cross-relation attention layer that distinguishes semantic differences among diverse behaviours (e.g., purchases vs. complaints) through relation-specific gating; 3) a reinforcement learning decision engine optimising marketing actions via Q-learning. Validated on a real-world e-commerce dataset (32,000 users, 142M interactions), DHGAT achieves an AUC of 0.892 in relationship prediction (5.7%-16.8% higher than baselines) and boosts marketing ROI by 41% in online A/B tests. The framework enables end-to-end optimisation of marketing strategies while balancing short-term conversions and long-term user value, offering a novel paradigm for data-driven marketing decision systems.

Keywords: DHGAT; time-decay edge weighting; cross-relation attention; brand-user relationship modelling.

DOI: 10.1504/IJICT.2025.149780

International Journal of Information and Communication Technology, 2025 Vol.26 No.39, pp.74 - 91

Received: 28 Jul 2025
Accepted: 15 Sep 2025
Published online: 12 Nov 2025 *

Title: Dynamic modelling of brand-user relationships via graph neural networks for enhanced marketing optimisation

Keep up-to-date