Research And Application Of Expert Information Knowledge Graph Based On Improved Canopy-Kmeans

In the new era,China continues to attach importance to scientific and technological innovation,and adheres to the development concept that science and technology are the primary productive forces.Therefore,various industries need to join a large number of excellent experts and researchers.The establishment of expert information treasure house and effective talent evaluation are conducive to facilitating expert information consultation,information retrieval and expanding academic cooperation in the field.To a certain extent,it can provide decision-making support for the government and promote the efficient utilization of talent resources in various fields,which has important practical significance.Therefore,this paper will build and apply the expert information knowledge graph based on the improved Canopy-Kmeans clustering algorithm.The main contents include:(1)The Canopy-Kmeans algorithm was improved and compared with K-means,Canopy clustering,DBSCAN clustering and Canopy-Kmeans clustering.The improved canopy-Kmeans algorithm takes the domain keywords of the supplemented results as clustering objects,generates the threshold of Canopy distance through cross verification,and calculates the result distance by using Levenshtein distance algorithm.The experiment found that the Canopy-Kmeans algorithm could greatly improve the accuracy of clustering results when it could obtain the optimal number of clusters,and the improved Canopy-Kmeans algorithm could also automatically extract the hidden information of results and generate keywords with high domain correlation of results under the premise that the accuracy of the improved Canopy-Kmeans algorithm was no lower than that of the canopy-Kmeans algorithm.More suitable for the current expert results database.(2)Combined with the improved Canopy-Kmeans clustering algorithm to generate expert fusion entities,the expert information knowledge graph based on fusion entities is constructed.Through the construction process of knowledge graph such as data acquisition,data processing,information extraction and knowledge fusion,a total of 178261 experts and their result entities and 197817 entity relationships were obtained and stored in the Neo4 j graph database.In the visualization part,CQL is used to query and display the part molecule graph.Compared with the expert information knowledge graph generated by the general method,the expert information knowledge graph based on the fusion entity can divide the expert results according to the research direction,and the display effect is more intuitive.(3)By combing and analyzing the scoring basis and characteristics of the existing basic indicators,h index and impact factor indicators,an expert scoring index system based on RFM is proposed to generate multidimensional expert scoring strategies.Taking the expert information knowledge graph as the data source,the score comparison experiment of the new index,h index and author influence factor is carried out.It is found that the score result of the new index is more flexible and takes more factors into consideration.Finally,the expert information retrieval tool is implemented with Flask framework to display the expert RFM ranking and achievement information,and the application of expert information knowledge graph based on fusion reality is completed.