Link prediction method based on local similarity

Abstract

The invention discloses a link prediction method based on local similarity. The link prediction method comprises the following specific steps: S1, obtaining original population data and constructing an initial population; s2, selecting a connection edge set with the same size as the test set positive sample from the node pair list without the connection edge as a test set negative sample; s3, obtaining community division corresponding to the population individuals by adopting an InfoMap algorithm; s4, obtaining a node pair list of the intra-community connection edge and the out-community connection edge according to the community division result; s5, starting from a first node pair in the network, according to the number of common neighbors of the node pair, sequentially calculating a common neighbor similarity index CN of each pair of nodes, and arranging all non-existing connecting edges in a descending order according to a CN value, so that the connecting edges arranged at the frontL pieces are most likely to have links. According to the method, the similarity index of the community structure information is added, so that the prediction accuracy is greatly improved.

Description

technical field [0001] The invention relates to a link prediction method, in particular to a link prediction method based on local similarity. Background technique [0002] Link prediction and community detection are two important directions in complex network research. Link prediction in the network refers to predicting the possibility that two nodes in the network that have not yet generated a connection may have a connection or will generate a connection based on information such as known nodes and connections between nodes. If entities and their relationships in society are abstracted into a network form—entities are nodes, and relationships are edges, then link prediction has great application value in it. The link prediction algorithm can identify unknown interactions, thereby reducing the cost of experiments. For example, in the biological field, most of the interactions between proteins are unknown, and researchers need to spend a lot of time and money to explore t...

AI Technical Summary

Problems solved by technology

[0005] In practical applications, it is difficult to obtain external information such as node attributes in most cases. For example, user information in most online systems is kept confidential.

In addition, it is impossible to determine whether the obtained information can truly reflect the situation of the node, and which information is useful to reflect the real situation of the node, that is, it is difficult to guarantee the reliability of the node attribute information

However, in methods based on link prediction such as network hierarchy, it is often necessary to generate many sample networks, so it can only deal with smaller-scale networks.

Images