Research On Internet Public Opinion Modeling And Evoluation Based On Bounded Confidence

The world is in a changing period, especially the emergence of social network,information is spread much faster than ever. Public opinion is spread rapidly through thevarious channels, but it also caused the spread of rumor and bad social contradictions, makethings become worse. So a very oractival research problem is how to guide the opinon to theright way.Since the information is carried by text, the first step in the analysis of public opinion isnumerical modeling of textual information, thereby extracting the relationship between thepublic opinions; secondly we need to model the relationship between users according to theinteraction between users; finally through evolution model to predict and analyze thedevelopment of events.Bounded confidence model considering the interaction between the users,studying theimpact of its intimacy, interaction rules, evolution rules, thresholds. Bounded confidencemodel shows its advantages in statistical physics,then introduced to the study of networkopinion, after years of research, several typical models come out. Hegselmann-Krause is oneof the best, mainly achieved good results in the simulation, but how to model the intimacybetween, threshold for a event in a real network research is still less studied.The main contribution of this thesis is as follows:1) introduces the principle of coroutine, and implement a distributed crawler frameworkbased on corotine. We use this crawler to get the information of tianya bbs, and introduce thedata update policy and information denoising mechanisms.We analyze the structure of theuser community, classify the user based on user activity and constructed user communitybased on the post-reply relationship, and model the users’ influence using PageRank.We givea information distribution model based on querying expansion, first when a keywordcomes,we expanse it by the keyword query expansion, and ultimately by the word frequencystatistics, building information model of public opinion.2) A history match method was propose based on particle swarm optimization, gettingthe parameters of a bounded confidence model by the history matching using PSO,experiments show that parameters got by history matching is better than using a fixedvalue,Finally we show our method result by example.3) An detail introduction to tizo opnion system. We give the detail introduction ofdatabase design, architecture design and the implement detail of each module.