Research On Opinion Leader Mining And Topic Discovery Technology In Microblog Precision Marketing

As an important product of the information development,micro-blog has rapid rise to the top with its characteristics of low threshold,real-time fame and so on,which makes it possible that users in different locations and in different time communicate with each other,and it has completed a metamorphosis from generation to popularization within a short time,then become an important public opinion platform Micro-blog marketing is taken more seriously by the merchant with the rise of micro-blog platform.Business enterprise image or product publicity should rely on micro-blog platform which needs to be spread by the opinion leaders,which often causes a new hot spot after all kinds of views 'collision.In the micro-blog marketing process,how to quantify the impact of the spread of opinion leaders?How to mine the opinion leaders who are appropriate for the business to make product promotion?How to find a topic which discussed by a large number of micro-blog users?All the above problems in micro-blog marketing should be solved.First,based on the existing work,according to the topic similarity,loyalty and positive degree of the fans to the user,this thesis proposed a method for mining opinion leaders based on user attributes and coverage.The model calculated the attribute value and the coverage area of users separately.By analyzing the timeliness of information and emotional tendency among users,we got the user set that have a higher content released by the user recognition and positive responsive the user,then we calculated the loyalty degree between users to get user' attribute ranking.At the same time,we used the topical similarity as weight to build contribution diagram and traversed each user' coverage by using breadth first algorithm whose limit depth is two,then we obtained each node's coverage by traversing.Finally,we got final ranking of opinion leader combining each node's coverage rate with attribute values.Experimental results show that the proposed scheme fits in with practical life in a certain degree,the opinion leaders' rank cut the star effect and improve the influence of micro-blog account whose contents are concerned about the people's livelihood.Next,this thesis presented the discovery topics model based on real-time co-occurrence network.Its main processes are:Structuring the word co-occurrence network,topic incremental clustering and topic ranking according to hot degree.First of all,we extracted the set of keywords from the primitive data and calculated the weight on the basis of the time parameter to structure the word co-occurrence network,and reduced sparsity by finding potential features of a strong correlation.Then,we achieved the topic incremental clustering by using the improved Single-Pass algoritum.Last,we tried to get the subject headings sorted by calculating their hot degree to get the most representative keywords of the topic.The above two research contents are both proved to be effective by experiments.We hope this method can bring practical help in the following aspects:1.Helping general public to understand social issues;2.Helping government to investigate the mood of the community;3.Helping enhance the awareness both products and corporates.