Research And Implementation On Extraction Technology Of Core Entities Based On Graph Processing In Non-public Sector Of Economy

Non-public economy is an important part of China’s economic system.Its population is large,widely distributed,diverse needs,and the relationship between economic entities is complex.Therefore,there is a high complexity in the monitoring and management of non-public economy.This paper studies the field of non-public economy,analyzes industry data by using graph calculation and other methods,constructs relevant algorithm models,adds core weights to enterprises in the industry,extracts core entities,and adds core entity dimensions to the enterprise user image of the National Federation of industry and Commerce for core entity enterprises,to assist the ideological guidance and policy support of the government or enterprise management units to enterprises.In view of the above research objectives,the main contributions of this paper are as follows:(1)This paper puts forward the definition of non-public economic core entity,expounds the non-public economic core entity from the two aspects of node attribute and network stability,puts forward that the entity core degree is related to the entity’s own attribute and position in the network,puts forward the definition of evaluation index of non-public economic core entity,and puts forward the method of extracting core entity.(2)According to the definition of non-public economic evaluation index,this paper uses the method of graph calculation to extract the core entity.For the specific data set of software and information service industry,the graph network is constructed from the perspective of investment and technology,and the PageRank algorithm is improved,inc luding weight reduction and iterative optimization.The final experiments show that the optimized algorithm has greatly improved the accuracy and convergence efficiency of the results.The graph network is constructed from the perspective of materials,and the graph convolution neural network method is used to make use of the previously marked core enterprise data to a certain extent and obtain high accuracy without adding new labels.(3)This paper proposes to take the industry as the dimension,extract the core entity in the industry,and realize the steps of data collection,data processing,network model construction and core weight addition.It also proposes the implementation method and overall process of automatic labeling for the core entity of the industry.This part will be used in the construction project of "online Federation of industry and commerce" of the National Federation of industry and commerce.