<p>Graph convolutional networks have achieved remarkable success in the field of multi-view learning due to their ability to effectively capture relationships and structural features among nodes. However, in real-world applications, labeled data are often scarce, which limits further improvements in model performance. Moreover, local structural discrepancies among different views and the inherent diversity of graph data pose additional difficulties for effectively modeling complex information. To address these issues, we propose a method named Label Propagation-based Multi-view Graph Convolutional Network for Semi-Supervised Classification. This method integrates multi-view graph construction strategies with a label propagation mechanism to mitigate the adverse effects of insufficient annotations, thereby improving classification performance under label-scarce conditions. Additionally, to further enhance the model’s performance, we introduce an edge weight adjustment mechanism, enabling the model to adaptively learn the relative importance of each edge during the label propagation process, thus improving both propagation accuracy and overall model robustness. Extensive experiments conducted on multiple public benchmark datasets demonstrate the effectiveness of our proposed approach.</p>