| A complex network is a network with characteristics such as self-similarity,selforganization,small world,scale-free.With the development of society,various fields can be abstracted into complex networks with the support of technology.Link prediction is one of the important methods for studying complex networks and understanding complex systems.It has indispensable value and significance in theory and application.First,in undirected and unweighted networks,in undirected and unweighted networks,most algorithms,such as algorithms based on node similarity and algorithms based on path similarity,only consider the node similarity or the path similarity,hybrid link-prediction algorithms for nodes and paths are fewer,and without considering the impact of the points.For the above reasons,this study proposes a hybrid algorithm of nodes and paths based on H-index(HPNC).By adding the point’s H-index value,it reflects the impact of the point in the network.Secondly,in undirected and unweighted networks,for the existing local similarity algorithms,most of them only consider the first-level neighbors and ignore the contribution of the second-level neighbors,and just describe the nodes with degrees.For the above reasons,this study proposes the local similarity of the local neighbors algorithm based on H-index(HLN).Add secondary neighbor node information on the basis of direct neighbors.The direct neighbor nodes and secondary neighbor nodes are different,Introduced the Hindex value to distinguish the contributions of different nodes.Introducing α to distinguish the different contributions of common neighbors and secondary neighbors.Finally,a hybrid algorithm of H-index-based nodes and paths and a local similarity algorithm of local neighbor nodes based on H-index were performed on many real datasets in different fields.The results show that both effectively improve accuracy.The HPNC algorithm verifies that the hybrid algorithm of node contribution and path is better than other algorithms in predicting results.The HLN algorithm verifies that considering second-order neighbor nodes and distinguishing node contributions is better than other algorithms based on node similarity. |
