Simulation network generation method applicable to estimation of network node classification method

Abstract

The invention discloses a simulation network generation method applicable to estimation of a network node classification method. The method comprises the following steps: firstly, generating a primary simulation network structure similar with a real network according to appointed topological structure parameters including the quantity of nodes, the quantity of edges, the maximum degree, distribution of the average degree, a small world, a community structure and the like; and secondly, generating a label of each node in the network according to appointed label characteristics, so as to obtain a simulation network with labels, which is applicable to estimation of a network node classifier. According to the simulation network generation method disclosed by the invention, a topological structure similar with the real network can be generated, and label information of the nodes is considered, so that the generated simulation network is relatively good for comprehensively estimating the node classification method; and in a network generation process, the topological structure and a label generation process are separated, so that the independence of parameter influences can be effectively guaranteed, and the dependence degrees to different attributes of the classification method can be more accurately estimated.

Description

technical field [0001] The invention relates to a simulation network generation method. Background technique [0002] As one of the important research fields of network science, network node classification technology has received extensive attention, and has important application value in the fields of identification, anti-terrorism, and information recommendation. The network node classification technology refers to: using the categories of some known nodes in the network to predict the categories of the remaining unknown nodes. Traditional classification techniques usually assume that the data are independent and identically distributed. However, there is often a strong correlation between network data, which makes the category of nodes not only related to their own attributes, but also closely related to network attributes such as neighbor nodes and topology. contact. Node classification methods can utilize the above features to improve classification performance. For ...

AI Technical Summary

Problems solved by technology

However, these labeled generation methods have fewer constraints on the topology, which makes the generated network topology far from the real network (for example, does not have a typical community structure, etc.), leading to the generation of such methods The network is also not suitable for evaluating classification methods

Images