Abstract: 【Purpose/significance】From the perspective of the user group,this study divides the community users into groups according
to user characteristics to understand the topic differences of different user groups and the community topics more comprehensively and clearly.Thereby,recommending academic resources for community users better.【Method/process】Using social network analysis and Topics method to divide the user group,then use the LDA model to carry out topic mining for different users.Finally,we use spectral clustering to achieve topic optimization.【Result/conclusion】The core users of the Science Network Information Science community have the same parts as the topics of general user groups,and there are also differences.The topics of the core user groups are more spe? cific and the topics of the general user groups are more extensive.Based on the virtual academic community user group topic mining model,the topics that community users care about can be displayed more comprehensively,and resources can be better recommended for community users.【Innovation/limitation】From the perspective of user groups,the study proposes a virtual academic community user group topic mining model to better recommend resources for community users.However,this study needs to be optimized in terms of da? ta,topic model and selection of social network analysis indicators.