Scholarlink

Studying the performance of graph neural networks (GNNs) using knowledge representation learning enhancement methods, we introduce alternating training of the knowledge representation learning module and the recommendation module to strengthen the GNN, prevent overfitting issues in recommendation algorithms, and optimize generalization capabilities. Using the Last.FM and MovieLens-20M datasets as experimental data, we analyze the KGENN algorithm based on two metrics: area under the curve (AUC) and recall. Practical results demonstrate that, compared to similar algorithm models, the GNN algorithm model studied in this paper exhibits higher performance and generalization capabilities.

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

Research on Generalization Ability of Graph Neural Networks Enhanced by Knowledge Graph

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Sujuan Zhao,
  • Jiali Xiao,
  • Fan Chen,
  • Kaiwen Yang

摘要

Studying the performance of graph neural networks (GNNs) using knowledge representation learning enhancement methods, we introduce alternating training of the knowledge representation learning module and the recommendation module to strengthen the GNN, prevent overfitting issues in recommendation algorithms, and optimize generalization capabilities. Using the Last.FM and MovieLens-20M datasets as experimental data, we analyze the KGENN algorithm based on two metrics: area under the curve (AUC) and recall. Practical results demonstrate that, compared to similar algorithm models, the GNN algorithm model studied in this paper exhibits higher performance and generalization capabilities.