Font Size: a A A

Research And Realization On Community Evolution Classification Method Based On Similarity And Difference

Posted on:2021-01-11Degree:MasterType:Thesis
Country:ChinaCandidate:S B YangFull Text:PDF
GTID:2370330602972060Subject:Computer Science and Technology
Abstract/Summary:PDF Full Text Request
With the rapid development of communication technology,mobile Internet grows explosively,the resulting dynamic network generated a large number of data,formed communities of different sizes and themes,mining the huge value contained in the communities has become one of the recent research hotspots.At present,the more common community research is divided into community discovery,community evolution analysis,community structure analysis,etc.The research on community discovery has made some achievements and has been applied to personalized recommendation and other fields.The analysis of community evolution is based on community discovery,further analyzing the changes of community and seeking the potential evolution law,so as to find the hidden deep level of information in the community,which provides the basis for predicting community evolution behavior,can also play an important role in public opinion analysis and control,protein function prediction,precision marketing,infectious disease transmission control,etc.This paper focuses on three aspects of community evolution analysis.A core node detection algorithm based on evaluation index is proposed.The algorithm is based on the evaluation index of node importance based on neighbor information and clustering coefficient,calculates the relative weight value of each node according to the evaluation index value,and determines the node with positive value as the core node to get the core node set.The algorithm does not require artificial threshold,and has great advantages compared with the algorithm based on node information and so on.A community evolution classification method based on similarity and difference is proposed.Based on the proposed algorithm of core node detection,the difference between core node and common node is determined by optimizing the difference formula.By combining the optimized difference formula with the existing similarity formula,A community evolution classification method based on similarity and difference named SDCE is proposed to determine the community evolution type from the similarity and difference.SDCE is compared with GED,SGCI on HEP-TH,Salon24,Facebook and other datasets.The results show that SDCE has a great advantage in detecting Forming and Dissolving events.It can avoid resolution limit,and the detection effect of small community is still good.At the same time,it has some advantages in the total number of evolutionary events,and the overall classification effect is better than other classification models.A dynamic community evolution type analysis system is designed and implemented.Based on the core node detection algorithm and the community evolution classification method based on similarity and difference,the system analyzes the basic information of the data set,and displays the evolution classification results in the forms of data,line graph,histogram,ring graph,etc.It improves the user experience,and users can get the information about data set and the classification results more intuitively.
Keywords/Search Tags:cluster coefficient, core node detection, the classification model of community evolution, dynamic community evolution type analysis system
PDF Full Text Request
Related items