This paper proposes a novel Heterogeneous Graph Attention Network (HAN) that addresses the challenges of graph neural networks (GNNs) in heterogeneous graphs, which contain multiple types of nodes and edges. The HAN incorporates both node-level and semantic-level attention mechanisms to learn the importance of nodes and meta-paths in a heterogeneous graph. The node-level attention focuses on the importance of meta-path based neighbors, while the semantic-level attention learns the importance of different meta-paths. By combining these two levels of attention, the HAN can generate node embeddings that capture the complex structure and rich semantics of heterogeneous graphs. The model is designed to be efficient, with linear complexity relative to the number of meta-path based node pairs, making it suitable for large-scale heterogeneous graphs. The HAN is evaluated on three real-world heterogeneous graphs, showing superior performance compared to state-of-the-art models and demonstrating good interpretability for graph analysis. The model's hierarchical attention mechanism allows for the selection of the most relevant meta-paths and neighbors, leading to more accurate and meaningful node embeddings. The results show that the HAN outperforms other methods in both node classification and clustering tasks, and the model's ability to capture the importance of nodes and meta-paths contributes to its effectiveness. The HAN also provides insights into the importance of different meta-paths and nodes, enhancing the interpretability of the model.This paper proposes a novel Heterogeneous Graph Attention Network (HAN) that addresses the challenges of graph neural networks (GNNs) in heterogeneous graphs, which contain multiple types of nodes and edges. The HAN incorporates both node-level and semantic-level attention mechanisms to learn the importance of nodes and meta-paths in a heterogeneous graph. The node-level attention focuses on the importance of meta-path based neighbors, while the semantic-level attention learns the importance of different meta-paths. By combining these two levels of attention, the HAN can generate node embeddings that capture the complex structure and rich semantics of heterogeneous graphs. The model is designed to be efficient, with linear complexity relative to the number of meta-path based node pairs, making it suitable for large-scale heterogeneous graphs. The HAN is evaluated on three real-world heterogeneous graphs, showing superior performance compared to state-of-the-art models and demonstrating good interpretability for graph analysis. The model's hierarchical attention mechanism allows for the selection of the most relevant meta-paths and neighbors, leading to more accurate and meaningful node embeddings. The results show that the HAN outperforms other methods in both node classification and clustering tasks, and the model's ability to capture the importance of nodes and meta-paths contributes to its effectiveness. The HAN also provides insights into the importance of different meta-paths and nodes, enhancing the interpretability of the model.