Research On Web-Analysis Based Pre-Warning Platform Of Mass Undergraduate Crisis

With the deepening of economic development, social development of our country has entered a period of transition. In this period, various contradictions of society get highlighted and exposed; the interests of the community are out of proportion and making rapid adjustment. It's easy to get mass undergraduate's thought and faith shocked and budged. Undergraduate students of the current generation possess strong self-awareness, but also have strong desires to participate in social affairs. However, if they can't express their appeal and desire, or get suppressed, it would easily lead the students to act irrational behavior, even to the occurrence of mass crisis, which would impact the normal order of colleges and our society. Colleges and universities are the window of our community, and students are the future of our country. Therefore, it is particularly import to effectively prevent mass undergraduate incidents from happening.The mass group incidents all have certain sense of organization, so there will be a lot of related public opinion information. And with the deepening and development of Internet, it has become a new kind of vector-borne for public opinion. In order to acquire the trend and situation of public opinion, it is intensively demanded that the construction of web-analysis based pre-warning platform for mass undergraduate crisis.This thesis takes BBS, Blog, News sites with commentary function as investigated subjects; propounds a scheme for mass graduate crisis pre-warning platform, which apply and integrate a variety of techniques such as public opinion information collection, content analysis, topic extraction, mechanism of early warning, etc. It also provides the useful ideas and exploration for analysis of public opinion and early warning of mass group incidents. Before making early warning, it takes processes of that analysis of public opinion, hot topic extraction, forecasting for trend of hot topic, pattern matching forecast results with sample data.The main results in this thesis include: the program design of the whole system; realization of subject-oriented focused crawler which bases on page similarity analysis and hyperlinks analysis, applying the properties of scale-free network in hyperlinks analysis; giving the realizable program of public opinion alarm subsystem; the realization of ARIMA based topic trends forecast module; proposing the design and realization of pre-warning generation module based on rules engine. All of the ideas provided above are helpful for the realization of mass undergraduate crisis early-warning platform.