HT-HAN: Hypergraph Attention Network Based on Hypergraph Transformation for Edge Classification

With the increasing number of graph data, traditional graph learning methods have encountered challenges in dealing with data with complex structure and high order relationships and in obtaining effective edge embeddings. In this paper, Hypergraph Transformation-based Hypergraph Attention Network (HT-HAN) is proposed to solve the edge classification problem in graph data. By introducing hypergraph transformation, we can obtain the hypergraph with the original edge as the node, and use the hypergraph attention mechanism to enhance the recognition ability of the model to the key information. Experimental results show that the proposed method achieves significant performance improvement on multiple data sets.