Research On Topic Discovery In Microblog Precision Marketing And Microblog Forwarding Prediction Technology

With the rapidly develop of the Internet,it has been become the main channel for people use the Internet to get information.Social media has been risen rapidly with the advantage of the Internet.It has become an online platform for people to communicate and share.Microblog has become the first choice for social media,just because of its fastly speed,simple operation,real-time sharing,and strong interaction.Currently popular Microblog market has become a new darling among merchants.However,the market method has stayed in the market mode centered on official accounts.Merchants publish advertisements on the form of Microblog.However,there are lots of problems such as misunderstandings,unclear user requirements,and unreasonable marketing tasks,the market results of merchants are often unsatisfactory.In view of the above problems,the Microblog precision market platform sets up an intelligent recommendation communicator module.It combines the characteristics of the Microblog platform and network marketing.This module mainly includes the research contents of Microblog topics detection and Microblog retweet prediction.But,due to the features of Microblog text is short and data sparseness,and it even includes issues affecting the forwarding of Microblog,Microblog market make it difficult for the intelligent recommendation of the communicators.For Microblog topics detection problem,the paper uses improved SOM neural network clustering.Firstly,for the Microblog feature sparse problem,the paper extracts the features of the Microblog text and then constructs the feature vectors.This method can be solved the problem of much high dimensionality in the clustering process.Finally,the method based on SOM clustering is used to topics detection.Experiments show that the overall index F value of the topics detection with the improved SOM neural network is improved by about 6.2% compared with the traditional topic discovery method.For the retweet prediction problem of Microblog,this paper analyzes the factors effect of the Microblog's retweet.The dissertation divides the factors of effect Microblog retweet forwarding into two categories: user attributes and text characteristics of Microblog.A method based on Logistic Regression model is proposed to realize the forwarding prediction of information.Through the analysis of user attributes and Microblog features,different Microblog features are used as input to the Logistic Regression model to predict the forwarding results.Experiments show that the overall index F value of this algorithm is on average about 4.5% higher than the traditional classification method.Finally,according to the above-mentioned theoretical method,the Microblog precision marketing platform is designed.The platform uses the intelligence to recommend the communicators for the merchants.The intelligent recommendation subsystem of the communicators is designed to visualize the platform.