Bibliometrics is a science to quantitatively analyze literature knowledge units by using mathematical and statistical methods and reveal the internal knowledge content of literature.Co-occurrence network analysis is a visual method to analyze the data relationship of document characteristic items in bibliometric research.According to the number of analyzed characteristic items,it can be divided into single co-occurrence network analysis and multiple co-occurrence network analysis.Compared with single co-occurrence network analysis,multi co-occurrence network analysis increases the dimension of feature items and presents literature knowledge more deeply.However,due to the increase of the dimension of the analyzed feature items,the number of nodes in the co-occurrence network increases,and the connection coincidence degree and crossover frequency between nodes are too large,which reduces the visualization effect of literature measurement.Therefore,at present,the bibliometric co-occurrence network analysis mainly focuses on single co-occurrence,and the visualization effect of multiple co-occurrence network analysis needs to be improved. 河北科技大学学报 2022年 第2期 翟君伟，等：基于LDA主题模型的文献特征项多重共现可视化方法 In order to solve the problems of too many nodes,too large connection density,disadvantage of discovering the value of data and low visualization effect in multi co-occurrence network,LDA topic model was introduced and the method of spatial division was adopted to transform the global visualization problem of feature items into subspace visualization problem.Firstly,the key words were extracted by using sati document title information analysis software,and the TF-IDF calculation was carried out.The calculation results were taken as the experimental data.Secondly,Python is used to construct a topic model for topic cluster analysis of the target literature set.Finally,Ucinet software was used to analyze the multiple co-occurrence of subspace documents with different topics,and the subspace analysis results are superimposed and reconstructed,so as to complete the structural expression of the multiple co-occurrence visualization system.The results show that compared with the original multi co-occurrence visualization method,the improved multi co-occurrence visualization method based on LDA topic model reduces the number of nodes in the co-occurrence network and the connection density between nodes due to the reduction of the scale of the multi co-occurrence network analysis system,that is,the number of documents and feature words in the subspace.It makes the structure of the multi co-occurrence visualization system clearer,increases the readability of the data,highlights the data value,and effectively improves the multi co-occurrence visualization effect.To a certain extent,this study can promote the in-depth research on knowledge mining of multiple combinations of literature constituent elements,and then improve the quality of empirical research on literature metrology in different fields.