نوع مقاله : مقاله پژوهشی
نویسندگان
1 دانشجوی دکتری هوش مصنوعی و رباتیکز، دانشکده هوش مصنوعی و علوم شناختی، دانشگاه جامع امام حسین (ع)، تهران، ایران
2 استادیار، دانشکده هوش مصنوعی و علوم شناختی، دانشگاه جامع امام حسین (ع)، تهران، ایران
چکیده
کلیدواژه‌ها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
This paper presents a simple yet effective approach to enhance the performance of Graph Neural Networks (GNNs) in node classification tasks. The proposed method involves incorporating the PageRank score—a global centrality metric—into node feature vectors to integrate broader contextual information beyond local neighborhoods. To evaluate the approach, three well-known GNN architectures—Graph Convolutional Networks (GCNs), Graph Attention Networks (GATs), and GraphSAGE—are tested on the Cora, CiteSeer, and PubMed citation network datasets. Model performance is assessed using standard metrics such as accuracy, precision, recall, F1 score, along with visual analyses based on Principal Component Analysis (PCA), t-distributed Stochastic Neighbor Embedding (t-SNE), and Uniform Manifold Approximation and Projection (UMAP). Experimental results demonstrate that adding the PageRank score leads to a significant improvement in classification accuracy, particularly in GATs, which benefit more from the additional global information. Despite its simplicity, the proposed method incurs minimal computational overhead and delivers consistent and reliable performance across datasets. Finally, the paper discusses the potential extension of this strategy through the integration of other centrality measures and its application to larger or heterogeneous graphs.
کلیدواژه‌ها [English]
)