Research On Correlation Analysis Based On Link Pattern And Community Detection Method With Random Walk With Bias

Complex networks are abstracted from various complex systems.The overall function of the network is reflected in the interaction among nodes in the network.Community structure is one of the most significant structural properties presented in many net-works.Generally,the community corresponds to the functional modules of the system.Th-erefore,extracting these communities of the network helps us to deeply explore the in-ternal rules in complex networks,it has important theoretical research significance and practical value for community detection of complex networks.As a result,it has been paid attention widely by many researchers,and many community-detection algorithms have been proposed,such as algorithms based on modularity optimization,label prop-agation,and random walk.On the basis of the full study of these algorithms,this paper proposes two community detection algorithms through the correlation analysis between the link pattern of vertices and the random walk process with bias.Community detection algorithm based on correlation analysis of link pattern.The algorithm first obtain the link pattern of the vertex,then calculates the correlation coef-ficient to obtain the correlation among the connection nodes and obtains the must-link and the cannot-link pairwise constraints.Expands the must-link and the cannot-link ac-cording to the transferability of the must-link.Then,acco-rding to the expanded cannot-link set cooperation seed node,the skeleton of the community structure is attracted to the must-link.Finally,the nodes that are not classified into the community is divided into corresponding communities by the method of minimum spanning tree,and the final community structure is obtained.A random walk algorithm based on the signal propagation mechanism with bias.The algorithm selects a node from the network as the signal source,choose the neighbor node randomly as the next hop node,and transmits the attenuated semaphore to the node,and iteratively randomly selects the next hop node and transmits the signal.Considering the attenuation of the signal,an attenuation factor is attached to each edge to constrain the signal propagation process.Through the propagation of the analog signal,each process of the network is repeated as a signal source to obtain a propagation matrix.Then,the self-loop is added for each vertex,and combine the similarity matrix with new attributes between the adj acency matrix and the similarity among vertices.Construct attributes for each vertex based on the new attribute matrix and propagation matrix.Finally,k-means algorithm is used for clustering to obtain high-quality community structure.In the end,k-means algorithm is used for clustering to obtain high-quality community structure with the minimum cost.In order to verify the performance of the proposed method,this paper conducts ex-periments on several real-world network datasets and synthetic network datasets,and compares the detection results with these of correlation algorithms.The experimental results show that the proposed algorithm can extract high-quality community structures from network.