Directed network link prediction method with fusion of multimode body information

Abstract

The invention discloses a directed network link prediction method with the fusion of the multimode body information. The method is characterized by comprising the following steps of: S1, constructingan initial network, and obtaining a node pair list without connection edges; S2, randomly selecting 10% of connecting edges in the initial network data as positive samples of a test set, taking the remaining 90% of connecting edges as a training set, and selecting a connecting edge set as large as the positive samples of the test set as negative samples of the test set; S3, obtaining role functionvalues corresponding to individuals in the initial network; S4, obtaining a role function Rw list of a common neighbor corresponding to each node pair; S5, obtaining the number of common neighbors ofthe node pairs; S6, obtaining an r'xy list of the node pair; S7, according to the r'xy list of the single die body, obtaining a list of rxy of the double die bodies in a superposition mode; or usinga machine learning method XGBoost to obtain a new score list according to the r'xy list obtained by different single die bodies. According to the method, the structural characteristics of the directednetwork are fully applied, so that the link prediction accuracy is greatly improved.

Description

technical field [0001] The invention relates to a link prediction method, in particular to a directional network link prediction method for merging multi-motif information. Background technique [0002] Link prediction is an important research direction in the field of complex networks. The basic problem to be dealt with is to predict the possibility of links between any two nodes in the network through known information such as network nodes and network structures. Through link prediction, we can not only obtain the possibility of future existence of edges that do not exist in the network, but also find out whether the existing edges in the network are false or missing. [0003] Among the link prediction methods based on network structure, the most commonly used method is the common neighbor similarity method. Liben-Nowellhe and Kleinberg found that the method based on common neighbors of nodes is one of the best methods for prediction accuracy. However, the common neighb...

AI Technical Summary

Problems solved by technology

However, in the link prediction of the existing directed network, the influence of the contribution value of the nodes other than the prediction edge of the motif on the link prediction is not considered.

Images