Scholarlink

<p>Trust-aware recommender systems are vital for enhancing recommendation precision and user satisfaction while addressing data sparsity and cold start challenges. Traditional approaches often emphasize limited aspects of trust or overlook the complex interactions between users and items. Moreover, knowledge-based methods may struggle with scalability, and artificial intelligence-driven models can miss uncertainties in item information and the multifaceted nature of trust. To address these shortcomings, we propose a hybrid recommendation framework named GUITARES. This framework employs graph attention networks to model enriched user-item interactions alongside trust relationships. GUITARES is structured into three main components: an ontology-based enrichment module that assigns confidence scores to item metadata, a trust prediction module that utilizes demographic similarities, and a graph attention module that integrates trust, metadata, and user ratings. GUITARES scales efficiently to large graphs—validated by systematic experiments—and achieves an RMSE of 0.80 on MovieLens, outperforming state-of-the-art methods, making it ideal for HPC environments handling massive recommendation networks.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

GUITARES: graph attention network for building knowledge graph-based trust-aware recommender systems

  • 期刊论文

  • 发表日期：

    2025年7月29日

查看全文
  • Ouissem Touameur,
  • Fouzi Harrag,
  • Ladjel Bellatreche

摘要

Trust-aware recommender systems are vital for enhancing recommendation precision and user satisfaction while addressing data sparsity and cold start challenges. Traditional approaches often emphasize limited aspects of trust or overlook the complex interactions between users and items. Moreover, knowledge-based methods may struggle with scalability, and artificial intelligence-driven models can miss uncertainties in item information and the multifaceted nature of trust. To address these shortcomings, we propose a hybrid recommendation framework named GUITARES. This framework employs graph attention networks to model enriched user-item interactions alongside trust relationships. GUITARES is structured into three main components: an ontology-based enrichment module that assigns confidence scores to item metadata, a trust prediction module that utilizes demographic similarities, and a graph attention module that integrates trust, metadata, and user ratings. GUITARES scales efficiently to large graphs—validated by systematic experiments—and achieves an RMSE of 0.80 on MovieLens, outperforming state-of-the-art methods, making it ideal for HPC environments handling massive recommendation networks.