| Link prediction aims to use existing network information to predict hidden or possible future links.With the in-depth exploration of this direction by researchers,link prediction research has achieved fruitful results.Many research results have been widely used in real-world scenarios,including friend recommendation,commercial marketing,network optimization,and protein function prediction.In theory,it provides new ideas for understanding the evolution of network structure.Therefore,link prediction research has important theoretical significance and practical value.In order to reduce the time complexity of the link prediction algorithm and improve the accuracy of the link prediction,this article uses a two-step strategy: First,the community detection algorithm based on greedy optimization technology(AGSO)is reasonably improved to make the community division more stable.Based on the community structure,reducing the link prediction range of the entire network to the community size can reduce the time complexity of the link prediction algorithm.Second,applying complex interactive behaviors to link prediction problems,by measuring the similarity of interactions between nodes,can effectively improve the accuracy of link prediction.Specifically,the main research contents of this article are as follows:(1)In view of the instability of the AGSO algorithm,this paper studies and proposes a community detection algorithm based on degree centrality local extension(DCLE).First,calculate the degree centrality of the nodes,take the sum of the degree centrality of the nodes at both ends of the link as the link centrality,and select the internal links of the community as the initial seed links.Secondly,local expansion based on greedy thoughts and rapid community division of the network laid the foundation for the following link prediction research.Finally,the experimental results show that the DCLE algorithm not only solves the instability problem but also can quickly and accurately identify the community structure in the network.(2)For the accuracy of link prediction,this paper studies and proposes a link prediction algorithm based on complex interaction behavior(CIBLP).First,the entire network is screened,the part of the network data with the closest interaction is selected,the interaction matrix is constructed and solved,and the weights of various interaction actions are obtained.Secondly,define the interactive similarity measure formula between nodes,and use the community structure as the search scope of node similarity to calculate the interactive similarity between unlinked nodes in the community.After that,the node interaction similarity of each community is summarized as the link prediction result of the entire network.Finally,the experimental results confirm that the CIBLP algorithm proposed in this paper can accurately predict links on social networks.(3)In order to make the CIBLP algorithm easier to apply to real scenes,this paper designs and implements an open platform for link prediction.By encapsulating the CIBLP algorithm securely,the platform provides portal management and link prediction function calls to third-party developers.Finally,the operation test results of the link prediction open platform show that the link prediction open platform has high reliability and security,which is helpful for the further promotion and use of the CIBLP algorithm. |
