Research On Opinion Community Discovery Technology In BBS

Network Forum (BBS), which contains a wealth of valuable information,consists of posts discussing different topics and has become an important opiniondissemination channel in today’s internet. In BBS, internet users involved in thereplies may form a series of communities spontaneously because of the sameopinions or views expressed in the topic. The existence and evolution of such acommunity structure plays an important role in mining the potential knowledge ofnetwork, monitoring and orientation of network public opinion information. Inrecent years, the network community discovery has been concerned by a growingnumber of researchers, with a significant research value.Based on the research findings of community discovery by domestic or foreignresearchers and scholars, and the shortage of the majority of the traditional researchon the foundation of the simple un-weighted network, this paper expands the targetnetwork for research into the weighted signed network and put forward the technicalline of this research. Firstly, combing the sociology triangle balance theory with thestatistical methods, this paper proposes a fast and efficient algorithm for mining theopposing opinion communities in the posts of BBS with the time complexityanalysis and an effective implementation. Secondly, through the introduction of theconcept of clusters-affinity, this pager proposes an algorithm based on thetechnology of hierarchical clustering to detect multiple opinion communities in theposts of BBS, gives its running process in detail and make analysis andimprovement at the level of implementation to deal with the bottleneck of thisalgorithm. Finally, we implement the two algorithms by programming and use theposts from BBS such as tianya.cn for experiment to test them. Then we analyze theexperimental results in depth. The experimental results show that in contrast, thepositive relationships in the same opinion community are much more intensive thanthey are in different opinion communities, and the negative relationships in the sameopinion community are much less intensive than they are in different opinion communities. This finding suggests that both of the two algorithms have the abilityto get implicit opinion communities in the BBS topics and have a strong practicality.