Research On Community Detection Algorithm Based On Network Local Information

Community detection in the network has received widespread attention because it can discover some useful information hidden in the network.Recognizing the community in the network also helps to understand and use the network more effectively.For example,detecting the community in the citation network can find relevant topics,and detection of shopping in the network community help build a recommendation system,etc.In today’s world,the real world is composed of various networks,so it has become a hot research issue to use the network community to find valuable information.Detecting communities in the network is an NP-hard problem.At present,many scholars have designed different community detection algorithms,and most of the existing community detection algorithms use the global information of the network,such as the network diameter,etc.However,when the network scale becomes larger,the difficulty of obtaining global information will also become greater,resulting in a decline in the efficiency of the algorithm.In addition,there are some algorithms that require prior information of the network,such as the number of communities.Through the analysis of the network,this paper proposes an adaptive community detection algorithm,which only uses the local information of the network to detect the community in the network.The algorithm first uses the local information of the network to calculate the importance of the nodes.Then according to the importance of the nodes,the seed nodes are selected in turn to construct the initial community,and the community is expanded.Finally,the obtained community is optimized.As the scale of the network becomes larger,the types and number of communities in the network are also increasing.However,users often only care about a certain community in the network,rather than all the communities in the network.Therefore,local community detection has gradually become a research.The local community detection is to find out the community structure where a given node is located.This paper designs a method for detecting local communities using semi-supervised learning of graph convolutional networks.The algorithm first designs the selection strategy of sample nodes.Then uses a two-layer graph convolutional network for training to obtain the membership degree of nodes to local community.Finally,set the membership threshold to filter out the nodes that meet the conditions.In the experimental part,the two algorithms proposed in this paper have carried out a large number of comparative experiments in the synthetic network and the real network.Through the analysis of the experimental results,the performance of the algorithm proposed in this paper is better than the comparison algorithm,and the algorithm is effective and robust.The performance is also better than the comparison algorithm.