Community Detection Of Multi-Attributed Networks Based On Stochastic Blockmodel And System Implementation

In people's daily life,complex systems play a very important role.There are also many different directions for the study of complex systems,such as the study of dynamics of complex systems,the study of randomness,and the study of selforganization and self-adaptability.This focus on community detection in complex networks.Community structure in complex networks is a very common and important feature of complex networks.It can enhance people's understanding of the characteristics of complex networks.It is of great significance to the study of individual behavior prediction and group division of complex networks.In recent years,with the arrival of the era of big data,researchers have found that node attribute information in complex networks can also provide many useful information for community to find problems.Presently,the community detection algorithm which integrates the attribute information of nodes has many disadvantages in the use of attribute information of nodes.For example,all attribute values of nodes are expressed in the form of 01 vectors.This method ignores the mutual exclusion constraints between different values of the same attribute,and combines all attributes linearly to change the attribute values of nodes into unit vectors in multi-dimensional space.This method can cause dimension disaster when there are more attributes.At the same time,it is found that the accuracy of this kind of algorithm in community discovery task is not ideal.In this paper,a generative community detection algorithm is proposed.This algorithm adds the parameters of the correlation matrix between attributes and communities based on the traditional stochastic block model.The correlation matrix is constructed by attributes,so that the independence between different attributes is preserved while the topological structure of the network and the attribute parameters of the nodes are fused.Then,the EM algorithm is used to get the best community division.In the process,in order to reduce the computational complexity,this paper uses BP algorithm to calculate the probability distribution of community partition after given attribute correlation matrix,which reduces the computational complexity of exponential level to linear level.After testing on different kinds of data sets,this paper verifies that the algorithm has higher accuracy.At the same time,this paper proposes a compatibility index between attributes and community partition,which can measure the contribution of attributes to community partition,help to interpret the results of community partition semantically,and also measure the aggregation of attributes or attribute combinations in complex networks,which provides a new idea for the semantic description of communities.Finally,a visual community discovery system is built using the proposed algorithm,which provides a simple example for the engineering application of community detection.