Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-11-29 Cooperative journals: 《计算机应用研究》
Abstract： In recent years, the detection of high-quality communities has become a hot topic in social network research. This paper proposes a game algorithm based on node attributes for community detection (G_NA) . This paper regards the process of community detection as the game of nodes in the network. When all nodes cannot improve their own utilities, the game ends. First, G_NA proposes a utility function based on the node degree attribute. Then, during the iteration, the nodes update the strategies in a sequence of node influence from large to small to increase their utilities. Finally, the proposed algorithm is compared with the existing algorithm in different real networks and artificial networks. The results show that the proposed algorithm is superior to other algorithms.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-24 Cooperative journals: 《计算机应用研究》
Abstract： The classical unsupervised clustering algorithm is fast, simple and爏uitable for mining large-scale datasets, and爄t燾an also directly divide communities. 燞owever, due to the complexity of communities, the classification accuracy of the algorithm is not ideal. Therefore, this paper proposes an error-correcting semi-supervised燾ommunity detection algorithm (ESCD) based on active learning. 營t can calculate the爐raditional k-means algorithm step by step, and adding pairs of constraints in the燾lustering process. In order to爌reserve爐he correct partitioning according to the爌rior爄nformation, we correct the wrong division to change the connection of the爊etwork. So爐hat爐he network has a more obvious block structure in the process of changing the distance between nodes and cluster centers. 燭he results of the experiment爏how that compared with the existing community discovery燼lgorithms, the ESCD algorithm has higher accuracy with爈ess爏upervisory information than other semi-supervised algorithms.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-02 Cooperative journals: 《计算机应用研究》
Abstract： In dynamic networks, detecting community structure is a very complex and meaningful process, which can better observe and analyze the evolution of the networks. For the community detection problem in dynamic weighted networks, this paper proposed an algorithm combining the community structure of the historical networks, called the evolutionary community discovery algorithm (ECDA) in dynamic weighted networks. The algorithm is divided into two steps: calculate the input matrix of the current timestep by combining the information of historical communities and network structure; and then calculate the result of community detection combining the historical timestep information through the input matrix. The algorithm has the following advantages: it can automatically discover the community structure of each timestep in the dynamic weighted network; the algorithm has a high sensitivity to the changes of network structure and the changes of community structure. And the experimental results show that the proposed algorithm can effectively detect the community structure in dynamic weighted networks, and it is quite competitive with other algorithms.
Hits
Hits
Downloads
Comment