Higher-order dependencies for multi-step link prediction

Multi-step link prediction methods offer significant potential across diverse domains, including trajectory prediction, recommender systems, and evolutionary game theory. By capturing higher-order dependencies among nodes, these methods enhance the accuracy of multi-step link prediction. In this paper, we introduce a novel algorithm for multi-step link prediction that explicitly considers higher-order dependencies within networks. To achieve precise multi-step link prediction, we propose a higher-order dependency network model based on flow data, selectively converting higher-order dependencies among nodes into higher-order nodes along with corresponding edges, and next devise an efficient algorithm. The effectiveness of our approach is demonstrated through empirical flow datasets, and we further apply it in the context of journal recommender systems.