Research On The Evolution Of Social Network Community Combined With Higher-Order Structural Features

With the development of the Internet,social media has become an important part of people’s lives,which also promotes the research and application of social networks.A social network is a type of network with diverse information and large scale.In the network,there is an interactive relationship composed of multiple nodes,which is called a high-order structure.In social networks,various relationships not only exist between two individuals,but also generally exist between multiple individuals,and present complex connection patterns.Existing social network research is usually based on low-level structures such as nodes and edges,and ignores the ubiquitous high-level structures composed of multiple nodes and multiple edges.Starting from the high-level structure,this thesis conducts noise reduction processing for a substantial amount of noise information in social networks and analyzes community evolution based on the noise reduction network.The main research work is as follows:(1)There are two main sources of noise in social networks.One is the false rumor information and robot accounts that exist in the network itself,and the other is the nodes and edges that are segmented during network slicing.Considering the ubiquitous high-order structure in the network,this thesis proposes a network noise reduction method combined with high-order structure based on the idea of spectral clustering.The purpose of network noise reduction is to identify and remove irrelevant or wrong nodes and edges,so the basic idea is the edge cut division in the graph partition problem.This study refers to the method of Benson et al.,which uses matrix eigenvalues to find the optimal partitioning scheme.Since graph partitioning is an NP-hard problem,there is no optimal solution,and the optimal noise reduction effect is difficult to define.Therefore,this study uses finite random walks for iterative solution,and introduces concepts related to the definition of spectral clustering and motif conductance.Experiments show that this method has good computational performance,and the noise reduction effect is better than the traditional method based on nodes and edges.(2)The evolution of online communities aims to study the evolution methods and mechanisms of communities,which has extensive application value and practical significance.Classical evolution models such as SGCI,PCET,etc.only consider node and edge information.Since communities usually evolve with multiple nodes and edges as units,this thesis introduces the high-order structure into the evolution model to extract more information.Rich information.In this thesis,the Jaccard similarity,node importance coefficient,and community scale indicators based on high-order structures are introduced into the GED model,and the judgment conditions of the model are improved.Experiments on the denoising network show that due to the reduction of noise information interference and the combination of high-order structural information,the model in this thesis has a better recognition effect in some evolution events and can obtain more reasonable evolution results.Experiments show that different types of data sets have different evolution event di stributions,and the development and changes of network communities can be predicted according to the event distribution.