| Division of complex network structure has penetrated into the research andapplication of various disciplines. The advantage of structurecomplex networktechnologyis that through the analysis of the network structure can find users, users withcommunity; community with community relations, these relations can really reflect theworld.This paper analyzes the complex network structure partition algorithm and scientificpapers recommendation algorithm. According to the structure of complex networkpartition algorithm and scientific paper recommendation algorithm in poor performance,they are analyzed and researched in this paper.Firstly, According to the structure of complex networks GN algorithm efficiency isnot very high, this paper put forward scientific network community structure partitionalgorithm. In this algorithm,the edge which is formed by the last deleted maximumbetweenness vertexand the other vertex is added the penalty value C, and recalculate theedge betweenness. Not all of the vertices is added to calculate the edge betweenness, butby successive addition of related vertices into the network structure.Secondly, for only relying on the module size to determine the final partition willcausepartition inaccuracy problems, error score was put forward and the combination ofthe both to evaluate the results of partition.Thirdly, the latent factor model of scientific paper recommendation algorithm israised to improve paper recommendation’s performance. Paper date andthe divisionnumber of the network structureare added in the algorithm. In order to enhance the abilityof real-time recommendation, first according to the user past history give a quickrecommendation listand thenupdate the result.Finally, the algorithm compared with no partition recommendation algorithm off-lineexperimental, and compared with other recommendation algorithm, like field, graph andco-author of academic relations algorithm to verify the effectiveness. |
PDF Full Text Request