Application Of Fusion PSO Algorithm And K-means Algorithm In Microblogging Public Opinion Monitoring

In the process of more and more extensive Internet coverage,social network has accumulated a large number of Internet users,which plays an important role in people's daily life.In particular,Microblog a social network product,is favored by a large number of netizens with its rich contents,brief text and rapid release and occupies the position of mainstream media.Microblog generates huge amounts data every day,which contains a huge amount of information.How to dig out the information and find the needed part has significant practical value and funct ion.This paper uses data mining technology to process Microblog data,find hot topics of public concern,track and monitor the online comments,so as to build a healthy network atmosphere and environment,obtain valuable information to provide the basis for commercial value,information spreading and sociological research,guide the online public opinion in a positive direction and provide more powerful guarantee to achieve the goal of a harmonious society.This paper took the data of Sina Microblog as the resea rch object.First of all,the data should be visualized so the distribution characteristics of which can be basically determined to carry out data preprocessing.As for the data of Sina Microblog is very complex and most of which are higher than three-dimensional,it is difficult to make an intuitive judgment.In this thesis,the k-means algorithm was used to cluster the data.However,the k-means algorithm is more vulnerable to the impact of the initial center point and usually an optimal solution be obtained in the process of data iteration.In order to solve these problems,the particle swarm optimization(PSO)algorithm was introduced in this thesis.The improved algorithm simplified the original algorithm and reduced its parameter settings,which can accelerate the convergence speed of the algorithm,effectively solve the impact of the initial clustering centers on particles and jump out the bondage of local optimum;the clustering effect of the algorithm can thus be improved.Finally,the clustering results of Microblog data mining were evaluated by using several different metrics.The evaluation index indicated that the clustering results of the improved algorithm were better than those of the traditional K-means algorithm.