.png){kind=icon}
.png){kind=icon}
Return
Simplicial motif predictor method for higher-order link prediction
.png){kind=icon}
PRE
AI
DOI：10.1016/j.eswa.2024.126284.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Higher-order link prediction is the task of forecasting complex connections or interactions beyond simple pairwise relationships in a network, offering insights into intricate network dynamics. Recent studies have delved into higher-order link prediction tasks within simplicial networks, which utilize simplices-higher-order constructs-to describe network structures. In this study, we introduce the innovative concept of simplicial motifs in the context of simplicial networks, and develop a supervised learning model that uses simplicial motifs as predictors for higher-order link prediction. Multiple empirical experiments demonstrate the superior performance of our simplicial motif predictor method (SMPM) compared to traditional metrics based on local/global information and other supervised learning methods. Furthermore, we observe a positive correlation between SMPM performance and both motif richness and richness distribution divergence. Based on these correlations, we propose two strategies for selecting simplicial motifs to optimize the efficiency of SMPM. Finally, in order to enhance the flexibility of SMPM, we explore the use of hyperparameter methods for prediction based on the above two strategies.
Keywords:
Simplicial network
Higher-order link prediction
Motif predictor
Journal
IF:
7.5
Papers:
2.9W
Citations:
10.2W