Research On Active Microblog Prediction Based On LDA And Random Forest Model

With the rapid development of mobile Internet,microblogging with its rich with the rapid development of mobile Internet,microblog is widely applied due to its abundant,timely information and pleasurable user experience.Microblog has become an important platform for people to exchange information and show themselves.Now,microblog has a large number of users that cover many social classes.Microblog users produce massive contents on the platform every day.The contents include not only the general public recording and sharing their daily life,but also the government,enterprises and other organizations making announcements about Social affairs.In general,it is the mainstream of microblog to watch and discuss hot social issues,which makes microblog an epitome of today's society.Given the huge social influence of microblog,the timely control of active parts on microblog plays an important role in the management of network public opinion.At present,prediction by microblog has become a research hotspot which covers a wide range.This paper studies prediction by microblog studied,including the mathematical model of microblog interaction and active microblog judgment,as well as the prediction model of active microblog.As for the mathematical model of microblog interaction,this paper first analyzes the three operations related to microblog interaction,and constructs the mathematical model of microblog interaction through weight solving.Then,this paper analyzes the correlation between the active value of microblog and the interaction value;and the correlation between the active value and the number of fans.Finally,this paper puts forward a mathematical model to decide whether a microblog is active.As for the prediction model of active microblog judgement,this paper designs the crawler and crawls the relevant microblog data based on the analysis of factors that influent microblog active value,and of the particularity of the research questions.Then,based on the pre-processed data set,this paper uses the active microblog judgment model to annotate the activity of microblog samples,so as to construct feature sets.After that,this paper uses LDA model to explore the theme of a microblog,acquiring the probability distribution matrix of document-topic via Gibbs sampling.The matrix is incorporated into the original feature set after discretization to obtain the optimized feature set.As for the algorithm selection for the prediction model,this paper analyzes a variety of prediction algorithms based on the data characteristics of microblog,and selects the random forest algorithm which highly fit with the data sets.Finally,this paper equalizes the random forest algorithm.In the end,on the basis of model training and parameter tuning,this paper uses contrast test to validate the effect of the prediction model on the microblog activity.Experimental results show that the prediction model of active microblog based on random forest algorithm and LDA theme model have good predictive effect.Therefore,this model is practical and feasible.