Design And Implemention Of Trending Topic Prediction System Based On Chinese Microblogging

With the increasing popularity of social network services, such as Facebook, Twitter and Sina Weibo, we are in the era of information explosion. The fast information sharing on these sites has made the information propagation more easily than ever before. So it’s important for business organizations and administrative decision makers to learn how to manage the risk of topics trending on these social network with fake information.In this thesis, we first formulate the problem of trending topic prediction on social network and propose a prediction mode based on SVM. Previous works on this domain primarily focus on time series analysis of posts, however, we divide topic data into time slices which is used as a unit of feature construction, and then assemble three subsets of features which supplement with each other to determine whether a topic will be trending on social network. To verify the performance of our prediction model, we design and implement a framework of crawler which is used to collected real data from Sina weibo.First, this thesis introduce the recent development of social network and then give a short description of research area of this domain. Second, a prediction model of trending topic on social network is proposed. The model is based on SVM and the input vector is assembled by three subsets of features. In the third section, the detailed design and implementation of the system is highlighted, including crawler framework, feature construction subsystem and prediction subsystem. The forth section is about the test work of the whole system, contain not only functional test but also performance test. In the end of this thesis, a short summary is given to show the whole work of the system.