Research On Collaborative Recommendation Of Virtual Academic Community Based On Researcher’s Dynamic Portraits

The development of network technology has broadened the way of communication and cooperation,making it possible for online communication and cooperation of scientific research.Researchers can share research results and participate in topic discussions by using virtual communities,which greatly promote their scientific research cooperation.The recommendation algorithm is to actively provide users with information services by analyzing users’ historical information,which effectively alleviate the current phenomena of information overload in the virtual academic community.The existing relevant researches focus more on recommendation of the papers and scientific research information,ignored the need of cooperation of scientific researchers.The advantage of user portraits is that can get the dynamic portraits of researchers,also conducive to deeply mining the cooperation needs of researchers,applied in the cooperation recommendation of virtual academic communities have a certain practical significance.First of all,in the previous two chapters,we had summarized the existing researches on user portraits and classic recommendation algorithms,analyzed the shortcomings of the existing research and propose solutions,briefly introduced the relevant theories involved in this study.Secondly,we had determined the data type,data source and construction process of the dynamic portrait in scientific researchers,clarified the label design principle of the dynamic portrait of scientific researchers,constructed the user natural attribute label,academic label and dynamic cooperation label system,and proposed the extraction and calculation methods of these labels.Thirdly,we find the key problems need to be solved in cooperative recommendation based on the dynamic model of researchers.By filtering characteristic topics to eliminate the influence of weak topics on research interests,adding the time function to build dynamic interests and calculate the similarity of dynamic research interests.Based on the transitivity of the cooperative relationship,we expand the cooperative relationship and calculate the value of the expanded relationship.We calculated the recommendation score by the weight fusion of relevant indicators.Finally,we obtained the data from “Baidu Academic”,a virtual academic community to verify the algorithm,and comparing with the content based recommendation algorithm,the user based collaborative filtering algorithm and the Mole Trust recommendation algorithm,the results show that the proposed algorithm performs better than the benchmark algorithm in terms of accuracy,recall and F1.