Multilayer heterogeneous network space node characterization method

Abstract

The invention provides a multilayer heterogeneous network space node characterization method, which comprises the following steps of S1, reconstructing a network according to a structure reconnectionrule, and generating a fusion graph based on a main layer and an auxiliary layer; S2, designing an algorithm of an adjacent distance, a structure distance and an attribute distance between node pairsfor the fusion graph; S3, in combination with the algorithm, obtaining the similarity between any two nodes in the fusion graph; S4, after similarity calculation of all node pairs in the fusion graphis completed, starting to construct a context network graph, and further generating an undirected unweighted graph; and S5, after carrying out random walk on the sampling path based on the context network, training a node representation vector through a skipgram model. Due to the facts that the multi-layer heterogeneous network is reconstructed, and the considered similarity is comprehensive, compared with other characterization algorithms, the method has a better effect.

Description

technical field [0001] The invention relates to a network space deep structure mining method under the condition of incomplete information in the field of cyberspace intelligent cognition, which realizes the identification of network hidden categories and potential categories, and specifically relates to a multi-layer heterogeneous network space node representation method, including Information fusion method of multi-layer heterogeneous network, general generation algorithm of network node representation, and network space node clustering method based on representation vector. Background technique [0002] At present, with the support of network information acquisition methods such as network traffic monitoring and active detection technology, the identification of cyberspace target organizational structure can be initially realized. However, first of all, the network structure itself is very complicated, and multiple local networks are often nested and intersected. Second, ...

AI Technical Summary

Problems solved by technology

[0007] The current technology mainly uses random walk plus skip-gram mode to train node representations. However, these methods mainly have two problems: (1) These methods do not consider the processing method of multi-layer graphs. If they are completely processed according to single-layer graphs, The accuracy rate will be greatly reduced; (2) Algorithms such as DeepWalk and node2vec only consider the adjacent similarity, and struc2vec only considers the structural similarity, and cannot comprehensively consider the adjacent similarity and structural similarity, and cannot take into account the attributes of the nodes themselves on the learned representation. influences

Images