Application Of Network Public Opinion Analysis In Early Warning Of Car Defects

Automobile industry is an important pillar of national development,and automobile product defects have potential adverse consequences such as great harm and wide social impact.Therefore,the effective supervision of the automobile industry is the key to protecting the rights and interests of consumers and promoting the healthy development of the industry.However,with the increasing of car ownership and the shortening of automobile R & D cycle,the current regulations on the recall management of defective automobile products in China are faced with regulatory difficulties,delayed response and other challenges.At the same time,in recent years,the market supervision plan and the government work report policy also put forward new requirements for the construction of "Internet+ regulation".On the other hand,online public opinion information has become an important basis for government departments and enterprises to make decisions.The research on risk early warning based on public opinion analysis has developed rapidly in many fields,and there have been early warning research on emergencies,new policy implementation,food safety,corporate reputation,public relations and product improvements,stock market volatility,etc.However,the research of public opinion analysis in auto defects early warning is just in the beginning,and there is no specific case in China.How to verify the feasibility and effect of public opinion analysis in auto defect early warning based on actual data is the focus of this thesis.This thesis takes small cars as the research object,crawls over 400000 pieces of unstructured effective data related to recall,complaint,public praise and basic parameters of small cars,applies entity recognition and improved emotional analysis method to deal with public opinion structurally,and solves the heterogeneous problem of public opinion from different sources with the help of the text similarity method proposed in this thesis.According to the experimental results of support vector machine and binomial logistic regression model,this thesis summarizes the early warning model of small car defects based on public opinion analysis,realizes the purpose of inputting the public opinion data related to the corresponding car subsystem,outputting whether the subsystem has defect risk and its probability.The experimental results show that the early warning of small car defects based on the analysis of public opinion in the network is feasible and effective.The prediction F1 value reaches 72.19%,and the accuracy rate reaches 94.75%.In addition,further analysis of the key monitoring indicators of the model can find that the early warning of small car defects is strongly related to the negative emotional intensity of public opinion and the origin,and there is a certain degree of "window breaking effect",that is,the car series and fault system with defects are more likely to occur again.In terms of the impact of recall events,recall can reduce a certain negative public opinion,but there is also a certain time lag.These conclusions provide an effective starting point for defect supervision.