| The vigorous development of socialized question and answer community(Q&A community)reshapes the ways of topic dissemination,knowledge sharing and emotional communication.The knowledge tag network with a certain structure will stimulate different topic popularity,and the topics with different popularity often reflect the emotional preferences of specific situations and groups.With the continuous transmission of information on Zhihu platform,the circulation degree of topics also changes with the horizontal and vertical time replacement,and the public’s attention to topic tags also changes accordingly.Thus,this paper takes topic tags as the breakthrough point and explores the topic popularity of knowledge network on Zhihu platform from different angles,which is conducive to the thorough understanding of the evolution rules for emergence and disappearance of social hot topics on Zhihu platform,easy to promptly understand the changes of public opinions on topics and make a move in time,controls the potential risk in advance and provides a theoretical foundation for topic guidance and public opinion control.Based on the neural network theory and time series analysis,this paper makes a multi-dimensional analysis of the topic data and its indicators in the Zhihu Q&A community,explores the general trend of change in popularity and obtains preliminary forecast results.Meanwhile,on this basis,it analyses the trend of topic emotion and explores the relationship between the development trend of topic popularity and emotional.Therefore,this paper mainly has the following research results: Firstly,the variance value of the popularity of each topic calculated based on the hierarchical definition method shows that the variance value of the popularity of entertainment news,epidemic and self-control is relatively large,while the variance value of the popularity of health,myopia,self-study and other topics is relatively small;Secondly,ARIMA(0,1,1),ARIMA(5,1,1)and ARIMA(8,1,1)are constructed for the three topics of "entertainment news","epidemic" and "self-control" respectively for prediction and fitting.The overall fitting degree between the actual emotional value of the model and the fitted emotional value curve is good,and all shows a gentle decline in the subsequent period.Thirdly,the emotional analysis API of Baidu has a high accuracy of 78% through the comparison analysis of emotional tools,and the time sequence diagram of topic emotional value and topic popularity,as well as the multi-variable LSTM prediction model constructed,all show that there is an inevitable relationship between topic popularity and emotional value.Based on the above research results,the hierarchical definition method based on the characteristics of topic data structure provides a new idea to study the topic popularity of online Q&A community.At the same time,in the study of topic popularity and emotional value,it can be seen that when a large number of Zhihu users under the topics expressed opinions or ideas produces a positive or negative impact,can play a role in promoting the popularity of the topic.Conversely,when the popularity of the topic is high,it can also attract other users to participate in topics,publish their own comments,express their emotional value,and play a role in the counter-promotion. |
