To address the challenge of inadequately capturing both semantic and structural features in heterogeneous networks,which often leads to unsatisfactory node classification performance,this paper proposes a novel method based on a hierarchical information aggregation graph convolutional network.The proposed framework employs a dual-branch architecture.In the meta-path aggregation branch,a hierarchical strategy is adopted.Specifically,a graph convolutional network combined with a hop-level attention mechanism generates single-hop representations,thereby enhancing task-relevant information while suppressing noise.Subsequently,a 1D convolution efficiently captures local semantic patterns from serialized meta-path information,followed by a meta-path-level attention mechanism to strengthen task-relevant meta-paths.Simultaneously,the neighborhood structure aggregation branch integrates local topological information from the heterogeneous graph.By collaboratively aggregating multi-level semantic and structural information,the model learns comprehensive node representations,leading to improved classification performance.Experimental results on ACM,DBLP,and IMDB datasets demonstrate the effectiveness of the proposed method.Furthermore,ablation studies verify the contribution of each individual component,offering a promising new approach for representation learning on heterogeneous networks.