HMNE: link prediction using hypergraph motifs and network embedding in social networks

Network embeddings, which map nodes to low-dimensional vectors, facilitate link prediction, a pivotal aspect of complex network research. However, existing methods often overlook the complexities of hypergraphs and potent structures for modeling intricate relationships among multiple entities. This paper delves into link prediction within hypergraph motifs and network embedding (HMNE), crucial for diverse fields like knowledge graphs and bioinformatics. HMNE employs motifs to perform network embedding, representing nodes as hyper-nodes. HMNE utilizes the skip-gram model to get the embedding vectors by analyzing the sequence generated using a local random walk technique. Additionally, we consider hyper-motifs as super-nodes to highlight structural similarities between nodes. To further refine our methodology, we use the depth and breadth motif random walk strategy on the embedded network with hyper-nodes. This innovative approach enriches our understanding of network dynamics and enhances the predictive power of our model. We have thoroughly experimented the proposed method on several real-world datasets, and the results consistently demonstrate its usefulness.