Design And Implementation Of Hot Topic Trend Prediction Based On Microblog Content

As a online social networking platform,Microblog has been developed as the most popular one on account that user can deliver message anywhere and anytime,and share information by clicking a like,commenting and reposting as well.In the process of development,the data traffic of Microblog could up to hundreds of millions of times per day,and the value of mining Microblog becomes increasingly significant.How to acquire and analyze data effectively and predict the trend of topics has been becoming an important research filed.In this dissetation,Sina Microblog data was used and the composition and transmission of the Microblog was analyzed.During the process of researching Microblog data,it is found that the processing of data is extremely difficult.The difficulty is that the release of Microblog information is arbitrary which means the user can publish information at any time.And the content of the Microblog information is short without the integrity of the article and the fragmentation is serious.Secondly,the speed of the transmission of Microblog is very fast.A highly concerned information will spread throughout the network in a few minutes,and it is difficult to predict the spread of information.It is a technical problem to discover the available information in time.Furthermore,How to effectively analysis the Microblog information dissemination and key users influence,for improving the system's prediction are very important.To solve these problems,the composition and structure of Microblog and the mode of information dissemination on the Microblog platform are analyzed,and a topic prediction system for Microblog content is proposed.The system is supported by crawling Sina micro-blog information data,taking big V and celebrity micro-blog data as a breakthrough point to study the key data nodes in the process of information transmission,screening out the effective data through the data filtering algorithm,using Bayesian algorithm to classify data and extracting data characteristics throughTF-IDF.Finally,through the analysis of the influence of the user's influence,the trend of the topic is obtained.The experimental results show that the effectiveness and accuracy of the system have reached the design requirements.