Predicting Future Intrablock Links in Directed Networks Using Triadic Patterns
Lekshmi S Nair, J. Swaminathan · IEEE Access · 2025
Complex networks model real-world relationships between entities where any two entities share more than one kind of relationship. Directed multilayer networks are used to represent such networks effectively, capturing the heterogeneity exhibited by the nodes and the directionality of relationships. The link prediction problem refers to predicting relationships (links) between the entities (nodes) that may arise in the future or identifying missing links to reconstruct the network. Our previous work proposed a novel block formation algorithm to determine future links between the blocks. This work proposes an intrablock link prediction technique within the blocks’ nodes exhibiting strong internal connectivity. The novelty of the proposed approach lies in extracting triads from a block using G-Trie by formulating a probabilistic model to predict the links between triads based on isomorphic properties, activity of nodes, and presence of common and influencer nodes. The experimental results on social, biological, and ecological datasets demonstrate that our hybrid framework achieved higher link prediction accuracy for both directed dense networks and sparse networks with lesser computational complexity.