| In recent years,the rapid development of computer and mobile internet technology has greatly changed the way people learn knowledge,and online question-andanswer community(Q&A)as a platform for knowledge and information sharing has developed rapidly.With the influx of a large number of users and data,the community is facing the problem of information overload,which makes users unable to screen knowledge and information quickly and efficiently,and reduces the user's personalized experience.For community managers,it is particularly important to keep abreast of the development trend of new technologies in online Q&A communities and the changes in users' personal preferences and interests.It can help the community to grasp the development trend of the current new technology in time and eliminate the unpopular questions,reduce the redundancy of the online Q&A platform and improve the user's search experience.Studying the mode and characteristics of user interest transfer in Q&A community can better depict the user's portrait characteristics,provide better personalized recommendation services to users,and improve the user's community activity and participation.Based on StackExchange online Q&A platform,this paper studies the development trend of tags in community from the perspective of tags,and analyses the transfer mode of user interest and the main influencing factors.The main research contents are as follows:1.Single source community tag popularity prediction based on machine learning: This study aims at the emerging tags in Q&A communities.Firstly,the popular and non-popular tag samples are labeled according to the frequency of usage.Secondly,the tag network is constructed according to the relationship of tags extracted from posts data of the community.Then,the network structure and statistical attributes of tags are extracted.Finally,the machine learning methods are used to construct the prediction model of tag popularity prediction.Experiment results showed that this model can predict the future trend of tags,and also verified the effectiveness of structural characteristics in the problem of popularity prediction.2.Multi-source community tag popularity prediction based on transfer learning: This research aims at the newly emerged community or the relatively cold community in the Q&A platform,to solve the problem that the tag popularity prediction model is ineffective because of the insufficient samples.This paper adopts the method of transfer learning,using data from larger communities to transfer knowledge to smaller communities.Due to the characteristics of network structure,the existing domain similarity measurement methods can't effectively evaluate the efficiency of transfer,resulting in negative transfer.Therefore,this paper proposes a new structural similarity measurement index,which has a strong positive correlation with single-source transfer efficiency.Then,a multi-source QA community tag popularity prediction model based on network structural similarity is proposed.Experiments show that this method can achieve better classification performance than the latest comparison methods.3.Analysis of user's technology interest transfer pattern based on Q&A tag: This study mainly analyzed the characteristics and influencing factors of technology community users' interest transfer model.In this paper,the community tag network is divided into communities by using the community detection algorithm.So,each tag is corresponds to the unique technology category.According to the user's historical technology category sequence data,an interest similarity measurement index is designed to characterize the user's interest change.The result of experiments show that in the technology community,the user interest change degree follows a power-law distribution,that is,most users have less interest transfer change;through the analysis of three factors: technology category heterogeneity,proximity and technology similarity,it is found that users' interests tend to transfer to similar technology field. |
PDF Full Text Request