Identification Of Enterprise Product Defects Based On Online Reviews

With the coming of the era of big data,various social networks such as We Chat,micro-blog,and Post Bar,all kinds of e-commerce shopping sites such as Tmall,Jingdong,Net Ease platform for the rise,to express their views and feelings using the Internet platform,sharing of information and re sources has become more common.In recent years,the e-commerce industry has brought vitality to the household electrical appliance industry,and the quality of household electrical appliances has been gradually highlighted.If the early detection of product defects,it could to minimize the damage caused by product defects to users and enterprise brands.Users will usually leave the evaluation on the Internet after the purchase of goods,product defects can be found from structured vast texts.At present Chinese product defect recognition model using supervised classification by tagging incomplete and low efficiency,and semi-supervised classification Tri-training algorithm performance is not good enough,to improve the product classification effect on Chinese network comment on the quality management of enterprise products have a certain significance.In order to analyze product defect content from the Internet environment user reviews,this paper through the analysis of models and methods made by their predecessors,based on the framework of product defect recognition model under the environment of Internet online reviews is constructed in this paper,including data preparation,text classification and text clustering.According to the characteristics of defect recognition of household appliances,the semi-supervised classification Co-forest algorithm is adopted to establish the comment classification model,and the defect identification of home appliance industry product reviews is carried out,and the product reviews containi ng defects are obtained.After the text classification,it contains the defect reviews,and then uses the BTM topic clustering algorithm to cluster the defect reviews,and gets the defect theme,the topic description details and the proportion.Through a br and of a best-selling dehumidifier,the above research carried out related experiments.The results show that the proposed Co-forest algorithm is used to identify defects with the existing supervised classification of product defect recognition method and Tri-training classification algorithm,Co-forest algorithm has higher accuracy and recall rate algorithm.Furthermore,the Co-forest algorithm identifies the post defect reviews,and uses topic based model BTM algorithm to cluster the topic,and obtains the defect theme results that can be understood by the product defect managers.In this paper,the domestic appliance industry product defects research,expanding the social media mining research scope,and can help enterprises in the production design,quality management and other aspects of the quality of feedback issues for rapid decision-making.