Research On Community Detection And Evolution Behavior Of Blockchain Digital Currency Transaction Network

With the continuous improvement of the influence and market value of cryptocurrency,researchers in different fields are paying more and more at-tention to the cryptocurrency transaction network.The modeling methods of cryptocurrency transaction networks still exist large optimization space.For example,the connections of the models remain to be explored and the char-acteristics to be distinguished.Specially,we are challenged to deal with the ever-increasing scale of cryptocurrency transaction data.And it is extremely important to construct a reasonable dynamic model to achieve a detailed anal-ysis of the evolution,and further to grasp the law of the development of cryp-tocurrency.This paper will perform research in the following three aspects to address the problems above.First,this paper explores the connections of three common static models of transaction network,and prove s the quantitative conversion relationship of the models.Because constructing static models is verbose and repetitive,we simplified the process of it.We propose one-mode projection of weighted di-rected bipartite graph,which is verified in the real Bitcoin transaction network.In conclusion,comparing the characteristics of three models,we recommend the "address-transaction" model as the static analysis model.Second,the models are challenged to perform analysis of dynamic evolu-tion behavior,dealing with a large number of transactions.This paper proposes a variable weight directed bipartite network model,which can dynamically show the state of network evolution through indicators such as the time-varying weight and node influence.We further introduce the mechanism of node dele-tion in Bitcoin transaction network.The mechanism changes the condition that the network nodes only increase in traditional analysis of transaction network,and reduce complexity of network analysis.Based on the index of information entropy of network we define,we propose the mechanism of screening the key evolution period to facilitate analysis of the evolution.In addition,the variable weight directed bipartite network model is used to analyze the static and dy-namic characteristic of the real Bitcoin transaction network from six aspects.And we find that node influence and transfer amount approximately obey the power-law distribution.Third,this paper further proposes the method of community detection of the transaction network based on the Louvain algorithm in the dynamic and static model,aiming at the shortcoming that the global model cannot reflect the local characteristics of the transaction network.We preprocess the naive Louvain algorithm,and introduce the pre-set parameter r to improve the flex-ibility of community detection.Further,we perform community detection of"global network" under the weighted bipartite graph model and the "effective network" under the VW-DBG model,and summarize three typical community modes.We confirm the existence of "core community" in the early Bitcoin transaction network.In conclusion,we perform community detection of the maximum component of the "effective network" in the evolution and study the details of the maximum component.We.analyze the proportion of "long chain"community in evolution.