Inferring links in directed complex networks through feed forward loop motifs

Abstract Complex networks are mathematical abstractions of real-world systems using sets nodes and edges representing the entities their interactions. Prediction unknown interactions in such is a problem interest biology, sociology, physics, engineering, etc. Most complex exhibit recurrence subnetworks, called network motifs. Within realm social science, link prediction (LP) models employed to model opinions, trust, privacy, rumor spreading media, academic corporate collaborations, liaisons among lawbreakers, human mobility resulting contagion. We present an LP metric based on motif directed networks, feed-forward loop (FFL). Unlike nearest neighbor-based metrics machine learning-based techniques that gauge likelihood node similarity, proposed approach leverages known dichotomy distribution networks. sparse, causing most associated links have low participation. Yet, due intrinsic motif-richness, few participate many distinct substructures. Thus, FFL-based combines presence absence motifs as signature outperform baseline ten biological datasets. conclude with future dynamic inference well its use designing combined varying orders features.