Abstract: Link prediction aims to mine unknown links based on observed topology information, which has high application value in many fields. At present, existing link prediction methods mainly focus on undirected network while the research of directed network is less. The prediction method based on structural information assumes that the more similar the nodes are, the more likely they are to be linked. Actually, the links between nodes are generated through paths, which cause similarity transfers between nodes. Most of the existing methods based on topology often focus on either the path between node pairs or the node degree, do not effectively mine the link length between node pairs and the local influence of nodes on the path. To solve this problem, this paper proposes a link prediction algorithm based on the effective connectivity path, and analyzes the contribution of different path length and node degree, semi-local centrality and H-index to node similarity. Compared with the existing eight prediction methods, the method proposed based on H-index effectively quantifies the local influence of nodes, and has a higher prediction accuracy under three indices.