Research On Multidimensional Expert Recommendation Based On Institutional Repository

With the rapid and convenient use of network resources,more and more information resources are continuously integrated into the Internet.Institutional repository is a way of information resource integration.As an emerging field,institutional repository has been continuously researched and developed in recent years,and more and more research has been done on the utilization of information resources in institutional repository.But at this stage,the research on the utilization of institutional repository is mainly aimed at the academic information resources in the institutional repository.The research on the experts who are the most important tacit knowledge carrier in the institutional repository is very few.What's more,at present,the demand for experts is reflected in various aspects,such as project review,paper review,competition review and so on,which require experts in related fields.Therefore,in order to make better use and mining the experts in the institutional repository,the personalized recommendation of the experts in the institutional repository is researched combining with the Chongqing Federation of Social Science Circles Institutional Repository.In this thesis,a multidimensional expert recommendation method of the institutional repository is proposed to meet the actual needs of institutional repository.This method uses Cascade-based hybrid recommendation algorithm.That is,using content-based recommendation method to make preliminary expert recommendation,and finally fusing collaborative filtering algorithm to produce the final recommendation results.Firstly,the method extracts the information of expert recommendation requirements of the institutional repository and the expert information of the institutional repository,and uses the TF-IDF(Term Frequency-Inverse Document Frequency)algorithm to construct the expert recommendation requirement vector model of the institutional repository and the expert information vector model of the institutional repository.Then,using the content-based recommendation method,the cosine similarity between the two vector models is calculated to generate a preliminary list of expert recommendations.Finaly,the collaborative filtering algorithm is combined,the users' interesting content by browsing the academic results published by other relevant experts and related experts,which is the basis of collaborative filtering recommendation.Combining with the multidimensional expert scoring model of the institutional repository to calculate the weight value of the corresponding experts,and then to adjust and filter the preliminary expert recommendation list.The top K experts ranked in the list are selected as the final recommendation result.In the process of research on personalized recommendation of experts in the institutional repository,a multidimensional expert scoring model of institutional repository based on PCA(Principle Component Analysis)is proposed for the situation that there are many dimensions of expert information in the institutional repository.This model constructs the multidimensional expert scoring information matrix of institutional repository according to the feature dimension of the expert in the institutional repository,and uses the PCA algorithm to calculate the dimensionality reduction of multidimensional expert scoring information matrix of institutional repository.On the basis of retaining the original main scoring information,the dimension that needs to be reduced is calculated,and a multidimensional expert scoring model of institutional repository is constructed.According to the expert data set in the Chongqing Federation of Social Science Circles Institutional Repository,the experimental analysis and verification show that the multidimensional expert recommendation method of the institutional repository proposed in this thesis can effectively carry out expert recommendation and meet the actual needs.