Heterogeneous Graph Neural Networks (HGNNs) have attracted significant research attention in recent years due to their ability to capture complex interactions among various node types in heterogeneous graphs (HGs). However, existing methods face critical challenges, including the loss of graph attribute information caused by excessive emphasis on semantic information and the difficulty of effectively integrating graph attributes with semantic information. To address these issues, this paper proposes HGNN-GAMS: a Heterogeneous Graph Neural Network for Graph Attribute Mining and Semantic Fusion. The model comprises two main components: graph attribute fusion and semantic aggregation. The graph attribute fusion module captures two intrinsic features of HGs—their unique topological structures and node attributes. The semantic aggregation module, leveraging an attention mechanism, integrates diverse semantic information within HGs. Ultimately, HGNN-GAMS fuses graph attribute features and semantic features to produce the final feature representation. This work pioneers the integration of graph attributes with semantic information and validates the model’s effectiveness through extensive experiments on real-world datasets.

With the strong capability of heterogeneous graphs in accurately modeling various types of nodes and their interactions, they have gradually become a research hotspot, promoting the rapid development of the field of heterogeneous graph neural networks (HGNNs). However, most existing HGNN models rely on meta−paths for feature extraction, which can only utilize part of the data from the graph for training and learning. This not only limits the data generalization ability of deep learning models but also affects the application effect of data−driven adaptive technologies. In response to this challenge, this study proposes a new model—heterogeneous graph neural network based on regional feature extraction (HGNN−BRFE). This model enhances performance through an “extraction−fusion” strategy in three key aspects: first, it efficiently extracts features of neighboring nodes of the same type according to specific regions; second, it effectively fuses information from different regions and hierarchical neighbors using attention mechanisms; third, it specially designs a process for feature extraction and fusion targeting heterogeneous type nodes, ensuring that the rich semantic and heterogeneity information within the heterogeneous graph is retained while maintaining the node’s own characteristics during the node embedding process to prevent the loss of its own features and potential over−smoothing issues. Experimental results show that HGNN−BRFE achieves a performance improvement of 1–3% over existing methods on classification tasks across multiple real−world datasets.

This paper constructs three bi-class multi-relational heterogeneous graphs based on real-world data, and it proposes a semantic and relation aware neural network model (SRA-BMHN) designed for bi-class multi-relational heterogeneous graphs. SRA-BMHN consists of two core modules: first, the semantic-aware module, which integrates diverse relational semantic information using a non-linear mapping function and attention mechanism tailored to specific relationship semantics; second, the relation-aware module, which employs a hierarchical bipartite subgraph aggregation strategy. This module first divides the bi-class multi-relational heterogeneous graphs into multiple bipartite subgraphs based on different edge relationship types and captures the topological information within them. It then merges these subgraphs into a relation-weighted subgraph and captures the relational features of nodes on the weighted heterogeneous subgraph. Finally, SRA-BMHN fuses the features from the semantic-aware module and the relation-aware module to generate the final node embeddings. Experimental results on three datasets demonstrate the superiority of the proposed method.