Information Feature Extraction And Policy Effect Evaluation Of Evidense-based Policy

The user-generated contents contain a lot of valuable information about the appeals,opinions,and sentiments of public,which was formulated by the discussions of public on social media,and it can be used as public policy evidence with the analysis of text mining.Firstly,with the user-generated content in social media as information source of evidence-based policy,a social media evidence-based policy making model was proposed with a modification of the agenda-setting theory and public sphere theory based on the characteristic of new medium era.Then an information feature extraction and policy effect evaluation method of evidense-based policy was constructed based on the mode: Analyze the network structure and user composition of social media platforms with social network analysis and demographic,to select the suitable social media platform as information source of policy making,which the community ambience is more consistent with policy development object;Evaluate the urgency of policy issue for the public using sentiment analysis,to decide whether it is necessary to formulate policies to solve this problem;Extract the major concerns and sentiments of public with a combination of text clustering,topic modeling and sentiment analysis,to formulate policies based on it;Simulate the public opinion feedback of proposed policy with a conjunction of counterfactual and sentiment analysis,to evaluate the policy effect and adjust it according to the results,so as to reduce the risk of policy mistakes.Finally,the validity of above methods was verified with the user discussion about haze topic in Zhihu Question & Answer Community as a case study.The analysis results show that:(1)the user-generated contents accumulated in social media platform contain the appeals,opinions and emotions of public,and the community ambience of social media platforms could be analyzed using social network analysis,to select the information source which is consistent with policy development object.(2)The major concerns and sentiments can be extracted from user discussion contents with text mining,which can be used as evidence to formulate policies that could meet the needs of public.(3)The public opinion feedback of proposed policy could be simulated with the counterfactual and sentiment analysis,to provide an evaluation result of policy effect.The above methods could provide valuable references for decision-makers,and improve the scientificity and rationality of policy making process.