Analysis Of Hotspots In Public Opinion Based On Events-Association

With the rapid development of information network technology, Internet, which is called the fourth media, brought almost subversive transformation to traditional media. In the Internet age, online public opinion gradually become the main form of social public opinion, It’s rapidity, extensity, in-depth, incisiveness and other advantages which traditional media can’t match, have severe impact on the mode of transmission of traditional media, and also a huge impact on the community. With the gradual deepening of network public opinion into people’s live, how to achieve effective network media guide evoked widespread thinking, the central clearly attaches, that we must pay attention to the great influence public opinion have on society, thus to analyze its development trend and make the right guidance program.To analyze the network public opinion, first we must have identification for hot events. In the majority of public opinion monitoring system, large number of duplicate reports describing the same event is all over for hotspot identification process. This is non-conductive either for viewing, analyzing, summarizing, or for the evaluation of events accurately, which requires the use of improved technology to perfect topic detection process. In this thesis, we propose an improved-single-pass algorithm to solve the problem of inaccurate calculation.Trend forecasting is the core of public opinion analysis, and the basis of public opinion guide. For trend forecasting, traditional algorithms have already yielded some results, but they do not have good prediction results in the primary phase of public opinion. This is caused by their too less consideration on the effect of the mutual relationship between events, for they are too simplified mathematical model that mainly meant for data themselves. In this thesis, we deal with this problem in a new way. Events that occur in real world are not isolating, there are some kind of links, which can influence their development mutually. Based on this analysis, an event-association based hot public opinion predication and analysis method is proposed. By analyzing the relationship between events, calculate the associated heat, and then use the gray prediction model, to get the correct prediction of the amount of information of public opinion trends.Innovations of the thesis are as follows:1. Improve the original single-pass algorithm to obtain a more accurate and efficient result.2. Heat prediction method based on event-association is proposed to get a good short-term prediction result on public opinion with small amount of information.