Reasearch On Community Discovery And Influence Maximization Based On Topic-attention Model

With the rise of some large social network sites,the research of the community discovery and the influence maximization based on social network has become one of the hot spots in the field.However,the current study is based on the social relationship to establish a social network as the mainstream,sharing topic and other social networks as the purpose is leading the trend of social network.In this paper,we study social network based on the topic,specific research contents are as follows.First,this paper construct a new social network model,topic-attention network,which is based on social relations and topic preference.In this network,the edges in the graph can reflect the explicit social relations among social individuals,it also reflects the social individual's participation in the network based on common interests.Secondly,the existing community discovery method based on the topic only regarding the topic as a property of a text or vertex,ignoring the status of the topic in the network topology,we adopt a new vertex similarity measure,which is defined by set pair relation,and considering the influence of topic and social relations,we propose a community partition algorithm CMTC based on topic-attention model.Thirdly,in view of the existing traditional network under the influence maximization method,lacking research on the transmission of a particular topic,ignoring the fact that the individual has different preferences for the information of different topics,and easy to be affected by individuals which is associated with them,In this paper,based on the set pair connection degree,greedy strategy is used to mine the influential vertexes,we propose an algorithm TA_CELF of seed vertex mining based on topic attention model.Finally,on the douban data set,analyze the impact of the topic on the community discovery through the number,size,distribution of the community and the topic of attention,etc.,verify experimental results of the algorithm CMTC;evaluate the experimental results by ISST,ISRT and ISRNT three indexes,verify experimental results of the algorithm TA_CELF.