Research On Community Structure Algorithm Of Decentralized E-commerce Network

In today’s era of mobile internet,various business channels and models are constantly changing,leading to changes in people’s usage habits and needs.As the saying goes,market demand determines the direction of the industry,and the decentralized e-commerce service industry is born from this.Decentralized e-commerce enterprises mainly rely on consumers’ private domain traffic to attract traffic customers.In order to better manage consumer community groups,consumer identification will be of great significance to them.In the decentralized e-commerce environment,consumers are often limited by social factors and differences in consumer demand,and the interaction between consumers is characterized by locality.Therefore,based on the locality of consumer networks,this paper proposes a decentralized e-commerce consumer network community division algorithm from the static and dynamic perspectives based on the Complex network theory,and further studies the relationship between its network power law characteristics and community structure,so as to achieve a deep understanding of its community structure evolution law.(1)A community partitioning algorithm based on improved LPA algorithm is proposed to address the shortcomings of label propagation distance and node similarity judgment in decentralized e-commerce consumer network community partitioning.First,aiming at the problem that the algorithm measures the distance of nodes according to the common relationship,the cosine similarity formula is introduced to measure the node similarity,and the similarity Adjacency matrix is constructed.In order to conform to the characteristics of locality and reduce the Time complexity of measurement,the principle of reverse lookup table is introduced.Secondly,the degree centrality index is used to select the initial center point,and the Clustering coefficient index is used as the label update rule.According to the label propagation characteristics in the algorithm do not meet the local requirements,the optimization formula of the label propagation distance is derived.The experimental conclusion shows that the improved LPA algorithm can effectively calculate consumer similarity,effectively limit label propagation,and achieve higher similarity between nodes within the community and lower similarity between communities,making it suitable for decentralized e-commerce consumer networks.(2)A novel incremental community partitioning algorithm based on the improved Dy Perm algorithm is proposed to address the shortcomings of incremental node identification and new node connectivity in decentralized e-commerce consumer network community partitioning.Firstly,based on the characteristics of local structural evolution,a rating matrix for each moment is constructed.The initial consumer network is constructed using the Pearson correlation coefficient formula.Considering the variability of individual consumer emotions,a rating matrix change mechanism is introduced,and the incremental node set is identified by comparing the rating matrices of adjacent moments.Secondly,according to the idea of preferential connection of new nodes,the consumer attraction factor is introduced,and the improved preferential connection formula of new nodes is introduced.Combining it with the mean field theory,the evolution dynamic equation of the Degree distribution of the consumer network can be derived.The experimental results show that the improved algorithm conforms to the evolution law of the consumer network,and can verify that the Degree distribution of the network obeys the power law distribution.The fitting curve of the Degree distribution conforms to the long tail theory.The formation mechanism of the community structure is positively correlated with the evolution of the power law characteristics of the Degree distribution.