Abstract: With the continuous growth of scientific and technological data,various science and technology departments have accumulated a large number of scientific and technological management data of scientific and technological projects.For a large amount of structured data,it is necessary to organize and analyze the distributed data,and finally provide data query and extraction ser-vices according to requirements.The analysis of relationships in relational databases is not effective.In order to improve the efficiency of analysis,relational graphs are introduced for data processing.Firstly,an entity search and localization algorithm based on word frequency is proposed,and the entities and relationships are extracted to construct the relationalgraph.Secondly,an improved FP-growth algorithm for frequent item mining of graph data is proposed in order to solve the frequent item screening problem in the graph data.Then,a data filtering process based on graph data is designed.In addition,this paper defines the scoring matrix,evaluate the screening data,and finally give an analysis opinion.The evaluation standard of data screening can be customized.Finally,combined with the constructed relational graph,the algorithm is applied in practice and encapsulated as a ser-vice.Experimentalresults show that the improved FP-growth-based frequent item mining algorithm has 10%~12% improvement over the traditional FP-growth algorithm.The accuracy of the data screening process designed in this paper reaches 97%.

Key words: Construction of human relation graph, Data analysis, Data mining, Graph construction, Relational graph, Service application