Design And Implementation Of Weibo User Interest Mining System Based On Improved LDA Model

With the continuous improvement of the level of informatization,the active users of Sina Weibo have continued to grow steadily,and at the same time massive amounts of Weibo data have been generated.Weibo data reflects users' interests and preferences,and the analysis of Weibo texts posted by users can help them discover topics of interest;at the same time,the platform can also make personalized recommendations based on user interests to achieve profitability.In summary,user interest mining is of great significance to enhance user experience and promote the healthy development of the platform.Usually Weibo documents have a long time span,and user interests are constantly changing,weibo texts are short and contain a lot of noisy data,which brings challenges to user interest mining.In order to effectively mine the interest and preference information of Weibo users,this thesis proposes a supervised dynamic topic model DS-LDA(Dynamic Supervised LDA),which is based on smaller time granularity.The model combines the time and label information of the document,and considers the topic-word distribution at different times.The thesis realizes supervised dynamic topic modeling;the thesis also compares this model with other topic models,and experiments show that it has achieved better results.In addition,in view of the characteristics of the Sina Weibo web page,this thesis also implements an easy-to-use user interest mining system:the data crawling module is responsible for crawling user microblog data,the data processing module performs stop-word removal and noise removal processing,the interest mining module predicts user interests based on the DS-LDA model,and the result visualization module displays the interest mining results.