Research On Key Technologies Of Internet Public Opinion Guidance

With the rapid development of microblog,wechat,Facebook,twitter and other new network media,online social network has become the main channel for people to obtain and disseminate information,accompanied by the frequent occurrence of various adverse public opinion events,so the research on network public opinion guidance is imperative.At present,there are three main problems in the guidance of network public opinion:(1)It is necessary to choose appropriate guidance strategies according to the development of public opinion situation,however,it is unrealistic to try various guiding strategies directly in the Internet because of its particularity,and the guidance strategies can only be chosen passively;(2)The existing effect evaluation methods of network public opinion can not solve the problem of evaluating the guidance effect quantitatively;(3)The existing network information diffusion prediction methods can not effectively predict the specific situation of a single piece of information diffusion,including the diffusion of information forwarding prediction and diffusion scale prediction,and the diffusion prediction of guidance posts needs to be improved.Key technologies of network public opinion guidance are studied in this thesis,and the main research results are listed as follows:(1)The choice of network public opinion guidance strategy is stuied,and a simulation deduction method of network public opinion guidance strategy based on communication theory is put forward.Firstly,the super network model is used to integrate the situation information of network public opinion events to form the network public opinion guidance simulation environment.Secondly,various guidance strategies are mapped into the network public opinion guidance simulation environment,and the dynamic simulation rules are designed by using the communication theory to simulate the network public opinion guidance strategy and output the proportion of negative public opinion.Finally,according to the proportion of negative public opinion,the appropriate strategy of network public opinion guidance is chosen adaptively,and the dilemma of trying various guidance strategies directly in the real network environment is solved.Experimental results show that the method is effective and helpful to the choice of network public opinion guidance strategy.(2)The evaluation of the effect of the guidance of the network public opinion is explored,and a method of the evaluation of the guidance effect of the network public opinion based on the fuzzy comprehensive evaluation is proposed.Firstly,a threat assessment index system of network public opinion is constructed,and AHP is used to determine the weights of each layer.Secondly,fuzzy sets of threat level are defined corresponding to situation indicators of network public opinion,and are furtherly integrated based on fuzzy-synthetical evaluation.Thirdly,the integrated fuzzy set of threat level is defuzzified,and the threat index of the network public opinion on the subject of the event is obtained.Finally,the guiding effect is evaluated by using the difference between the two threat indexes of before and after network public opinion guidance.Experimental results show that this method can accurately evaluate the effect of public opinion guidance,which is completely consistent with the experience judgment.(3)The propagation prediction of network public opinion guidance posts is researched,and an information diffusion prediction method based on feature attenuation enhanced neural network is presented.Firstly,the multi-level neighbor influence mechanism is introduced to distinguish the influence weights of different levels of neighbors,and the user feature representation is updated combined with the network representation method.Secondly,combined with the time attenuation effect,the feature attenuation enhanced neural network model is constructed.Finally,the information forwarding prediction and information popularity prediction are carried out by using the constructed model.Experimental results show that this method outperforms the mainstream diffusion prediction method based on end-to-end neural network,and it is helpful to predict the influence range of guide posts.