The Research On StackExchange Q&A Network Data Mining

StackExchange is one of the most popular Q&A websites,including more than 100 Q&A communities,which plays a very important role in the process of knowledge transmission.Our purpose in this thesis is to study the complex relationship between users,Q&A,and tags in the Q&A communities of StackExchange,by utilizing the popular complex network methods.Combined with the geographical information of users,it will help us better understand the pattern of knowledge transmission in the Q&A communities.In chapter one of this thesis,we first introduce several popular Q&A websites,and review important studies in this area.Then,in chapter two,we introduce complex network theory,including the structural characteristics of complex networks,classical complex network models based on them: ER random network model,WS small-world network model,BA scale-free network model,and information dissemination models.In chapter three,we introduce the dataset of StackExchange,and the concept of task-oriented social network.We construct user-question and question-tag bipartite networks,and then build Q-A relationship networks between users and tag similarity networks between tags by projecting these bipartite networks.We then visualize and characterize these networks.Moreover,in chapter four,we select four Q&A communities in StackExchange,and study the evolution of their tag similarity networks with time.We find that the knowledge structure of Q&A communities of different type has different evolving patterns.To study the pattern of knowledge transmission in the Q&A communities of StackExchange in United States,in chapter five,we first construct Q&A relationship network model for the states of US.Although this model can help to identify that California and New York are two relatively important states in the process of knowledge transmission,it cannot help to clearly determine the knowledge transmission source and pattern.To solve this problem,we propose a novel knowledge transmission network model,based on which we can exactly find the knowledge transmission sources.More than one knowledge transmission sources can be found in most Q&A websites in StackExchange,and the knowledge in Q&A communities can be transmitted from more than one sources to other states as time evolves.Besides,we also compare the structural characteristics of Q&A relationship networks and knowledge transmission networks.We find that the connections between states are relatively dense in Q&A relationship networks,while their structural patterns are relatively different across different Q&A websites.However,in knowledge transmission networks,the connections between states are sparser and the average shortest path lengths are relatively longer,making the transmission skeleton much clearer.Meanwhile,the Q&A websites have more similar knowledge transmission patterns in this case.These results indicate that,compared with the Q&A relationship network model,our model can better explain the knowledge transmission patterns in the Q&A communities.Finally,in chapter six,we conclude our work and look forward to future work.