Network alignment based on multiple hypernetwork attributes

The network alignment problem refers to how to find the node correspondence across different networks in multiplex networks. This study has significant implications in various disciplinary fields. However, current network alignment work focuses on simple networks. These methods based on simple networks are doomed to fail to capture high-order relationships. In order to fill the gap in this area, this paper will introduce a prediction method of inter-layer connectivity based on multi hypernetwork structure attributes. Among them, the hyperedge similarity index of nodes is specially designed for higher-order relationships in hypernetworks, and a degree punishment mechanism is designed to reasonably evaluate the similarity of higher-order relationships between nodes in different environments. This method further considers the quantity and strength information of the similarity of higher-order relations, which helps to further increase the accuracy rate on the hypernetwork. We compare this method with other advanced methods on different real-world hypernetworks and artificial hypernetworks. Experiments show that the method has good performance and robustness. In the biological metabolic network, the accuracy of this method can even be improved by 29.8% compared with the comparison method. In the 38 groups of data we tested, the accuracy rate was 78.9% when it was stably higher than other methods, 15.9% when it was not lower than the comparison method, and only 5.2% when the method performed poorly.