Research On Scholar Collaborative Recommendation Algorithm Based On Knowledge Graph

With the rapid growth of online academic resources and the increasing number of researchers,many scholars can not find the right scientific research collaborator with yourself efficiently and accurately.In order to solve this problem,this paper makes an in-depth study on personalized recommendation in the field of scholar cooperation.Considering that knowledge map can fully reflect the semantic association between entities and the advantages of deep learning technology in information representation,this paper proposes a personalized recommendation method based on deep learning technology,combining knowledge map and scholar cooperation behavior sequence,focusing on the construction of knowledge map of scholars' cooperation field,the vectorization representation of entities in the knowledge map,and The recommendation technology of academic resources based on knowledge map.The main work of this paper is as followsFirstly,the knowledge map based on the background of scholars' cooperation is constructed,and the knowledge map is applied in the field of scholars' cooperation.By extracting the internal entities and relationships,the knowledge map is stored in neo4 j database,and finally the final knowledge map of scholar cooperation relationship is formed;Secondly,making full use of the advantages of Vectorization in recommendation technology,the classical translation model Trans E and its derivative models are used to deal with the vectorization of knowledge map.According to the comparative experiments of different models,the Trans D model is finally determined as the best knowledge representation learning method of the recommended model in this paper;Then,the behavior sequence data set of scholars' cooperation field is obtained.On the basis of aminer academic resource data set,the number of scholars' cooperation and corresponding years are added,and the corresponding preprocessing operation is carried out to realize the data set matching to obtain the final behavior sequence data set of scholars' cooperation field;Finally,a recommendation model based on knowledge graph and scholars' cooperation behavior sequence(RKSC)is proposed.The model uses knowledge representation learning method of knowledge map to obtain the structural features of academic resources,and uses word2 vec And word vector sequence average pooling technology is used to capture the unstructured features(text features)of academic resources,and then the attention mechanism is used to fuse the above two features to obtain the final vectorized representation of academic resources.Then,GRU neural network is used to capture the characteristics of scholars' cooperative behavior in the current sequence,and the long-term preferences of scholars are obtained The characteristics of sub cooperation represent users' short-term preference.The short-term and long-term preference features are combined to obtain the potential cooperation interest of the scholar.Finally,another scholar who meets the current preference is recommended for the scholar.As far as this paper knows,this is one of the few attempts to combine knowledge mapping with deep learning technology,to face the academic community and consider the long-term and short-term preferences of scholars,and to use it in the field of serialization recommendation.In addition,the loss function in the model is improved.Finally,the experimental results show that the proposed RKSC recommendation model has better performance than the related work.