Research And Application Of Community-based Collaborator Recommendation Model

With the rapid development of Internet technology,the amount of information has increased dramatically,and researchers will spend more time looking for suitable partners,which limits the academic innovation and development of researchers.If we can recommend suitable partners for researchers,which can effectively promote the exchange and innovation of knowledge,save time for researchers to find collaborators,and ultimately improve the scientific research ability of researchers and improve the output of scientific research results.The content of the community-based collaborator recommendation method proposed in this thesis mainly includes the following parts:Firstly,to deal with the randomness of the existing semi-synchronous label propagation algorithm in the label propagation process,a site-effect-based semi-synchronous label propagation algorithm(SESLPA)for community detection is proposed.The algorithm defines the influence coefficient according to the edge clustering coefficient and its extension,assigns different influence coefficients to the different order loops formed by nodes within three degrees based on the Three Degrees of Influence Rule(TDIR),and the edge influence can be calculated to expedite label propagation and make the community structure visible.Results showed that SE-SLPA improves modularity performance and normalized mutual information,reduces the number of iterations,and enhances the stability of the algorithm.This research is vital to the study of social networks.Secondly,a discovery model for authors' interest evolution(wBATF)based on biterm and academic ability is proposed.The existing model obtains the evolution trend of the authors' interest based on the abstract of the documents,ignoring the feature sparseness of the short texts,and does not distinguish all the authors of the document,resulting in inaccurate distribution of the author's topic.The wBATF generates topic-sharing bitrems extended the features of the short texts,in addition,quantifies the research achievements of authors,and then assigns academic ability.In the web of science datasets the Gibbs Sampling is adopted to estimate its parameters and verify the validity of the model.By analyzing the experimental results,the model can improve the accuracy of the author's topic flow evolution trend and reduce the perplexity of the model.Thirdly,a community-based collaborator recommendation method(CCR)is proposed.When calculating the similarities of scientific researchers,the weight of time is introduced,and the closer the scientific research results are,the greater the influence is on the similarity among researchers.The process of data acquisition,processing,community detection,topic discovery and collaborator recommendation is designed and implemented,and the algorithm is verified.that it provides a new and effective way for collaborators to recommend.Finally,the community-based collaborator recommendation algorithm is applied to design and implementation of the researchers' collaborator recommendation system,including overall design,module design,key technology implementation,etc.,and verify the practicality and effectiveness of the community-based collaborator recommendation method.