Research And Implementation Of Skyline Technology For Public Opinion Oriented Analysis

Public opinion is an important method to realize and study the variation of social hot-topic, early warning of the emergency and leading the social public opinion. Microblogs is a new kind of information release and social network platform which become very popular in recent years. Microblogs public opinion is the union of attitude, belief, comment and emotion expressed by people through this information media. Microblogs have deeply affected our human daily life and the analysis requirement for microblogs public opinion is of great practical significance. Therefore, the research of microblogs public opinion analysis attract extensive attention.The research of skyline query is one of the order-sensitive query techniques and got lots of attention after 21 st century. Compared with other order-sensitive query techniques, it wants to provide some valuable objects for users’ selecting later instead of an evaluation function which should be given by users, so that it can processing some cases which top-k query techniques cannot. For the characteristics of network environment changing rapidly and social opinion topic suddenly happening, skyline query technique have a more better adaptivity. Therefore, the research on microblogs public opinion analysis based on skyline query technology is quite valuable.In detail, we highlight the main work of this thesis as follow:1. We defined the skyline query problem for multi-dimensional data set according to the multi-dimensional feature of microblogs dataset. The problem is mining the dominance-consistent pairs in the object data set. For this problem, we proposed a theoretical model and detailed it from the mathematical aspect, including theorem proving and complexity analysis.2. We analized the connection between our problem and a frequent partial order mining problem. We proposed a basic algorithm to solve our problem based on the Frecpo algorithm for the frequent partial order mining problem.3. In order to optimize the performance, we proposed two algorithm for the dominance-consistent pair mining problem according to the microblogs data set feature of data attachment and sequential access. Then we tested our algorithm in a famous synthetic data set. The result certified that the algorithm performance has obvious improvement.4. We implemented above algorithm on the existing public opinion analysis system, including two modules: user influential spreading analysis module and key-case finding module. The experiment results on the real data set gathered by the public opinion analysis system suggest that our work’s effectiveness.