Research On Evolution And Trend Prediction Of Public Opinion At Medical Insurance Fraud

Nowadays,China has built the world’s largest medical insurance system,and the coverage has stabilized at more than 95%,greatly alleviating the burden of medical treatment for people.However,some people illegally raised medical insurance funds by various means,seriously threatening the stable development of the medical insurance fund.As medical insurance is related to everyone’s vital interests,once such incidents burst out,it will lead to a heated discussion among the people,which will easily form a network sensation crisis and affect the stable development of society.Therefore,it is necessary to conduct an in-depth analysis of medical insurance fraud incident,extract the main ideas and demands from public opinion,analyze the public opinion evolutionary process,and finally predict the public opinion development trend.This paper studies the evolution analysis and trend prediction of public opinion at medical insurance fraud.This paper selects the event of "Hospital employs healthy people to cheat medical insurance" as the research object.This paper is developed from the perspective of public opinion object,and Word2 Vec is applied to judge the object of each comment.In evolutionary research,we first divide the public opinion propagation stage according to the life cycle theory,then use LDA method to mine the subject of comments and identify the emotional focus.At the same time,we adopt Snow NLP to conduct emotional tendency research,analyze the public’s attitude towards different objects,so as to express the public’s feelings and emotions.We use quantitative analysis to study the public opinion object,time sequence information,comment subject and emotional tendency in the public opinion evolution process.On this basis,this paper constructs features from the public opinion object,time sequence,comment theme and emotion dimensions.After feature selection,this paper utilizes and compares many mainstream prediction algorithms to build a forecasting model.Finally,the optimal feature and algorithm combination of public opinion trend prediction model are obtained.The results indicate that:(1)The evolutionary research results of public opinion theme and emotional tendency are highly consistent,that is,people turn from discussing the event itself to considering the essence and consequences of the event,and finally put forward hopes and suggestions,which are consistent with the theory of cognitive-affective consistency;(2)By testing and comparing LSTM,linear regression,support vector regression machine,decision regression tree,ridge regression,back propagation neural network and other mainstream prediction algorithms,it is found that RMSE,MAE and R2 of LSTM are all optimal.that is,LSTM model is more suitable for predicting the development trend of public opinion;(3)The public opinion trends prediction model based on multi-dimensional characteristics of public opinion evolution proposed in this paper is superior to the model only using traditional time series dimension characteristics in terms of overall indicators and response speed of public opinion outbreak.The results of this study can provide a powerful reference for the relevant departments to listen to the public opinion,guide public opinion’s development direction,and early-warning the public opinion crisis,which could provide a powerful reference for improving the relevant laws and regulations of medical insurance to ensure the stable operation of medical insurance system.