The Application Of Collective Intelligence In Social Media

Social media has become a new platform for people to share, communicate and interact with others. There forms a variety of groups according to the different interest, topics. The groups will produces very large amount of information by daily communication, which has the characteristics of disorderly and short periodicity. How to obtain high quality information to provide users with more complete information and service is paid more attention to., There exists and reflects the wisdom of crowds in the groups, therefore, using the collective intelligence to analyze the information of the social groups is more important for servicing the group users.Aiming at the application of collective intelligence in social groups, this paper uses the collective intelligence to solve the problems of the spam recognition and the sorting algorithm, to provide convenience for users in the group.First, with the amount of spam increased in the microblog plat, using the micro topics in sina weibo as corpus, by combining the wisdom of crowds and the unsupervised learning, this paper proposes a method based on the Random Walk. It is an automatic spam recognition model, this model constructs a network by computing the similarity between microblogs, through the clustering algorithm of Random Walk model to find groups. Finally, it works on the individuals and small groups by using the dominant characters of spam to recognize the spam. Experiments show that the model can effectively recognize the spam microblogs, and is better than the traditional supervised learning method, especially in the aspect of the recall value. It could reduce spam and convenient users to browse the microblog.Second, this paper also works on the movie group in sina weibo, and proposed a ranking model based on the collective intelligence which exploited the Ant Colony Algorithm to make the rank on the information of micro-group. At the same time, it also combined the extent of user’s preference and the hot extent of movies to give a comprehensive ranking. Besides, the model considers the emotion factor to analysis and compute the users’ emotion. The final ranking is based on the accumulative value of emotion about the hot movies in the group. The experiment shows that this model can meet the users’ preference better than other method. And it has some degree of real-time properties to provide users with relevant information of movies effectively.