Defect Discovery Method Of Vehicle Based On Social Media

With the rise of social media,a lot of valuable content were generated every day.Some topical content may have an unimaginable influence compared to traditional media.In China,more and more enterprises began to use social media to promote corporate image or corporate products,interact with consumers,and maintain consumer loyalty to enterprises.However,there was almost no gap in how to use the data from social media to mining product defects to improve the level of defect management.So far,in order to reduce the product defect rate,a large number of quality management methods had been proposed.However,after the product putting on the market,a variety of unexpected defects could not be found.In the field of automobile manufacturing,there are two main ways for companies to discover automobile defects in advance:simulation test before delivery and after-sales customer survey.In the age of social media,any unexpected flaws may spread on social media,which will cause great damage to the reputation and production of the company.In order to reduce the damage of accidents to enterprises and improve the level of defect management,this paper expanded the specific defect categories and proposed an improved social media-based vehicle defect recognition frame according to the actual needs of enterprises.An improved feature selection method was proposed and the three feature extraction methods were compared in detail based on the naive Bayesian classification method.In addition,the EM algorithm was used to realize the semi-supervised learning of defect recognition.The experimental results showed that the improved feature selection method can significantly improve the accuracy of classification when the defect was classified probably.The advantage was obvious when the data size was large;the semi-supervised learning method combined EM algorithm can achieve the same defect recognition effect when the data with label is only a half compared with the supervised learning.Therefore,the framework proposed in this paper can significantly reduce the time and labor cost of enterprise and provide decision support for enterprise defect management.