Research On The Method Of Network Public Opinion Analysis And Threat Assessment For Emergency

With the rapid development of computer and the abroad popularization of Internet, peopleexpress attitudes and opinions about the various social issues by exchanging on the Internet. Thenetwork public opinion is not only people’s true reflection of the social event expressed on theInternet but also concentrated embodiment of the social public opinion. While the network publicopinion spreads with a rapid speed, affects widely and jeopardizes greatly, which make analyzingand monitoring of the network public opinion have an important significance.The information of the network public opinion is more vulnerable to be concerned becauseof its complex background, huge scale and various styles, especially for the network publicopinion of emergency. How to analyze and control the trend of the network public opinion timelyand effectively is a problem to be solved urgently. Therefore, to satisfy the requirement ofanalysis and threat assessment on the public opinion, an index system of the network publicopinion is constructed, which is using to analyze the important index of the public opinion,furthermore, the model of the network public opinion threat assessment is also designed.The main contents and achievements of this thesis are as follows:1. An index system of the network public opinion for emergency is constructed. For thesystematic shortage of the index system, this index system, which contains four major categories:the emergency, the public opinion network user, the public opinion media, the public opiniontrend and sixteen indexes in detail, is constructed from the three aspects of the public opinion:origin, development and result. Indexes of the public opinion are quantified by some methods. Ahybrid endowing weight method, which is used to determine the weight of the index of thenetwork public opinion, is proposed. Finally the rationality of the index system is analyzed fromthree aspects: harmony, redundancy and relevancy.2. A new sentiment orientation analysis method, which is based on the sentiment word andthe semantic pattern, is proposed. For the shortage of the sentiment orientation analysis of thenetwork user, the sentiment lexicon and the semantic pattern, which have a correlation with thetopic review of the network public opinion for emergency, are constructed. A new sentimentorientation analysis method is proposed, which increases the influence of the semantic patternbased on analyzing the sentiment orientation of the word, to realize the polarity of the networkpublic opinion. The experiment shows that this method improves the accuracy rate of judging thesentiment orientation greatly.3. A threat assessment model of the network public opinion based on the BP neural networkis designed. To offer samples for the BP neural network, a sample training method of judging thedegree of the threat based on the fuzzy comprehensive evaluation, is proposed, which relies on the fuzzyness characteristic of the public opinion index and the facility to fuzzy judgement of thefuzzy comprehensive evaluation. To satisfy the need of the threat assessment model of thenetwork public opinion processing automatically, a threat assessment model of the networkpublic opinion based on the BP neural network is designed, which has the advantages ofhandling the incomplete data and finding the inherent law between the input and output. Finallyusing the Matlab to train and test the model by inputting the obtained samples, and thesimulation experiment shows the validity of the model.