Research On Community Partition Technology Based On Multiple Attributes

With the advent of the Internet era and the rapid development of information technology,a variety of large-scale online network were generated.The emergence of these networks promoted the research of complex networks.From the protein relationship networks,transportation networks and scientific papers cooperation networks to the micro-blog networks,commodity recommendation networks,online movie commentary networks,human's understanding of complex networks gradually deepened.Community structure,as an important aspect of complex networks structure analysis,embodies the similarities and differences among individuals in the complex system.The study of community partition helps people understand the structure and evolution of complex networks more clearly,and provides techniques and methods for the macro-control and analysis of the network.This paper gives an overview of the classic algorithm and evaluation method for community detection,and focuses on the research of the community partition about the social networks with multiple attributes.At present,the numbers and types of attributes associated with nodes in social networks are rapidly increasing.Some community partitions more depend on the joint information of multiple attributes of nodes to achieve the goal of effective partition,such as food culture,watching movies' tastes and other abstract communities.The existing multi-attributes networks community partition mainly quantifiy the node's multi-attributes as the node's attribute vector,and then uses the traditional community partition algorithms on the basis of computing the similarity between the nodes.There are two shortages in this process.One is that the distance calculation on simple attribute quantification can not express the close degree of attribute adequately.Second,the different effects of attributes on the results of partition are not effectively reflected.In this paper,we put forward a method to define the attribute similarity function,which is based on the characteristics of individual attributes.Meanwhile,useing the method of optimizing parameters give each attribute a weight coefficient to reflect its importance,so as to improve the similarity measure ability of multi-attributes nodes.In addition,combining random walk community partition algorithm,the similarity between nodes is regarded as the possibility of random walk.Thus the similarity matrix of multi-attribute network is transformed into the transition probability matrix of random walk,and use the random walk algorithm to obtain information expansion matrix between nodes for community partition.Aiming at the problem of slow community merging in this algorithm,this paper uses the optimization strategy of multi-group merge simultaneously to process information diffusion matrix in order to reduce the dimension of the information diffusion matrix and improve the speed of community merging.Finally,the proposed multi-attribute commu nity classification technique is applied in the film evaluation network to find the similarity movie taste of among users.Experimental results show that this technique can improve the performance of community partition effectively.