Research And Application Of Relevance Analysis Model Based On Heterogeneous Information Network Embedding

With the rapid development of informatization construction,a large amount of data has been accumulated in the field of scientific research.Behind the rich scientific research management data are hidden information between colleges,disciplines,and researchers.The related information can provide scientific guidance for scientific researchers on project cooperation and other scientific research activities.How to mine and effectively use this information is a problem that needs to be solved urgently.In recent years,network embedding technology has developed rapidly and is widely used in data mining tasks.After expressing scientific research data in the form of a heterogeneous information network,the use of network embedding technology can realize the mapping of the original network to the lowdimensional feature space,which is convenient for inference learning of the network.Most of the existing embedding methods for heterogeneous information networks focus more on structural information,ignoring the additional auxiliary information in the network,which will impair the quality of embedding and thus affect the accuracy of the relevance analysis results.In view of the above problems,this paper uses GATNE model as the theoretical basis for network embedding to carry out in-depth research,and proposes a relevance analysis model based on heterogeneous information network embedding.This model supplements the node-level attention mechanism during the process of learning the node embedding of aggregated neighborhood information,effectively distinguishing the importance of each neighbor node;at the same time,it also merges the interaction items represented by the node and the neighborhood from similar neighbors to aggregate more information.After the embedding vectors of each node are obtained,the degree of relevance between the nodes is measured by calculating the cosine similarity of the vectors.Subsequently,the model was evaluated on the scientific research data set,which verified that the model could improve the accuracy of the relevance analysis results and reduce irrelevant objects in the retrieval results.Finally,this paper applies the relevance analysis model to the scientific research management platform,and plans the application's organizational structure and module design in detail.Based on the relevance measurement results obtained by the model,it realizes the functions of retrieving related objects and displaying the related basis,and perfects the scientific research management,has good research and application value.