MoTIvE: a motif-based framework for node ranking in temporal networks

Abstract

Identifying influential spreaders in temporal networks is essential for understanding and controlling information and disease propagation. Among existing centrality methods, local methods have gained considerable attention due to their computational efficiency and their ability to identify influential spreaders using neighborhood-based features. However, most local methods primarily rely on degree-based or local path-based information, overlooking higher-order local structural patterns among interacting nodes. Higher-order structural patterns provide valuable insights into the underlying interaction dynamics and influence propagation behavior within temporal networks. To address the limitation of existing local methods, the paper introduces a novel method, MoTIvE , which leverages temporal motifs as local structural features to characterize higher-order, time-respecting interaction patterns among small groups of nodes. Temporal motifs encode complex temporal dependencies while preserving computational efficiency. The proposed framework first computes three-node temporal motif counts for each node and then incorporates motif information from immediate neighbors to enrich the node’s structural representation. Finally, the motif counts of a node and its immediate neighbors are aggregated to derive an overall influence score for identifying influential spreaders in temporal networks. Experimental results on nine real-world temporal networks show that MoTIvE outperforms existing methods in accurately identifying influential spreaders.