Multi Dimensional Recommendation Of Online Community Resources From Global Perspective

[Purpose/significance] The network community provides information services to users through personalized resource recommendation system,in order to solve the problem that resources can not be reasonably allocated and efficiently spread.However,the existing resource recommendation system can't find the real relationship between resources deeply and accurately,and it can't find the user's interests accurately.It can't realize the multi-dimensional mining of user's interest resources,and it mostly realizes the same kind of resource recommendation or one-dimensional resource recommendation.In the long run,the "information cocoon room" effect is prominent,not only the types of information resources accepted by users are gradually single,and the scope of information is gradually narrowed,but also the selective psychology of users to obtain information is affected,and the comprehensive development of users' cognition is restricted.Therefore,it is necessary to implement multi-dimensional recommendation of online community resources.[Method/Process] This study innovatively proposes a multi-dimensional knowledge association system of network community based on super network technology and network representation learning method based on shallow neural network from the global perspective,and realizes multi-dimensional recommendation of network community resources based on this system.Firstly,this research divides resources into three types: user,text and word resources,and regards them as the combination of user,text and word granularity knowledge units,which transforms the problem of multi-dimensional resource multi relationship discovery into the problem of multi association mining representation of multi granularity knowledge units.Secondly,all granular knowledge units and their multi-layer heterogeneous networks are described in one network by using the super network technology.The super network of network community resources is constructed to retain the information of knowledge units and ensure the overall situation.Then,the line algorithm based on shallow neural network is used to deal with the super network uniformly.In the unified feature space,the multi granularity knowledge units are represented as low dimensional vectors.Thus,the multi knowledge association system of network community is constructed,and the association type and association strength between multi granularity knowledge units are revealed deeply,clearly and accurately.Finally,based on the multi knowledge association system,the semantic representation of user interest is realized accurately,and then the multi-dimensional mining of user explicit interest resources,user potential interest resources and user dynamic interest resources is carried out.Combined with certain visualization technology,the multi-dimensional recommendation of interest resources is realized.[Result/Conclusion] Through the empirical analysis in dingxiangyuan cardiovascular forum,the scheme achieves multi-dimensional and high-precision recommendation of interest resources for target users,and verifies its feasibility and effectiveness.This not only effectively alleviates the "information cocoon room" effect,but also optimizes the quality of information service in the network community,which helps the network community to highlight its own advantages,enhance the stickiness of existing users,attract new users,and realize the long-term operation of the community.