Research On The Support Prediction Of Public Policy Sentiment Based On Social Media

With the arrival of the Web2.0 era,the widespread information technology has great influence on human's idea and behavior.From the perspective of network public opinion,with the advancement of the process of social democratization in China,the awareness of the people's participation in political participation is increasing,and the cognitive tendency of the Internet users for the current policy can feed back a large number of hidden information and promote the formulation and evaluation of public policies in China.However,the current research on public policy prediction in China is mainly focused on three aspects: the exploration of policy evaluation index system,the trend of public opinion evolution,and the analysis of topic views,there is fewer research on the development of public opinion of the policy inclination.This paper constructs a predictive model of public policy opinion based on social media.It regards the change of the public attitude of policy as a random process and uses microblog online comment text as experimental data to explore the future tendency of a certain policy.Specifically,the model consists of two parts: In the first step,we build a decision support model of public opinion and generate socialized parameters of Markov prediction model,avoiding the error caused by subjective parameter setting in previous studies.The model firstly constructs the policy goal,object,plan,and effect evaluation of the four dimensions,from the perspective of sentiment analysis,we use the domain framework semantic dictionary and policy comment topic dictionary to identify the policy dimensions of microblogging and construct a single text policy evaluation vector,Finally we use two methods of weight measurement to synthesize the public's final public opinion support policy.The second step is to build a Markov forecasting model and update the original public opinion support based on the social influence of users.The state transition matrix is solved by combining genetic algorithm with the quadratic programming algorithm.The prediction accuracy is corrected by the dynamic error compensation formula,which is used Matlab to predict the iterative equilibrium point when the system reaches a balance.In order to verify the effectiveness of the model,the paper finally selects the established policy of "Delayed Retirement policy" as an example to make an empirical study.The experimental results show that: first of all,the prediction model proposed in this paper can effectively predict the future tendency of Internet sentiment for people through the comparison of realistic policies;Secondly,compared with the solution of single genetic algorithm,the Mark off matrix solution method proposed in the prediction stage can improve the accuracy of experimental prediction.Finally,through case analysis,it can be known that the public has a low proportion of non-objectionary policies on delayed retirement policies and there is a lot of policy risks.The government needs to continuously coordinate with the various tasks of policy prior assessment to ensure its smooth implementation,and at the same time,In order to conceive and regulate the supervision of social media in China,this article puts forward a number of suggestions and hopes to make some contribution to the research on public opinion in China's policy areas.