Research On Product Defect Recognition Model And Algorithm Improvement For Social Commerce

With the rapid development of the internet economy,product defects will have a fatal negative impact on enterprises.Timely and accurate detection of product defects can minimize the damage caused by product defects to users and corporate brands.Social commerce features a combination of user interaction and e-commerce on social media,and its rich user-generated content provides important references for enterprises to improve product quality.How to discover and quantify potential product defects from a large amount of user-generated content is the focus of scholars and companies.Firstly,on the basis of scholars' research,the existing social media analysis framework(SMART)is improved,and a general framework for product defect identification is constructed by combining different platform features.Secondly,the PESI model is constructed based on platform,enterprises,society and individual features.And then,according to the outlier detection theory,SD_DCOF method is proposed to identify spam comments and ensure the validity of the comments.Thirdly,after comprehensive analysis and comparison,information gain(IG)is used as the feature evaluation function,and SVM is used to identify the defects of different platform data,and the product defect corpus is constructed to effectively improve the accuracy and recall rate of recognition.Finally,the LDA model is improved by combining product component and internal product defect information.The defect subject clustering is adopted to identify the product quality problems and provide the enterprise with decision-making information on product quality management.This paper extends the enterprise-centered internal management of product quality to the user-centered external quality control that identifies product defects from comments,applies big data analysis and text mining to product quality management,and deepens and expands the related research of product quality control.This study has important theoretical value and practical significance.The main research contents and innovations of this paper are as follows:(1)A general framework of product defect identification for social commerce is proposed.Based on the improved SMART framework,a generic product defect identification framework is constructed with four stages: data preparation,spam comment identification,product defect identification,and product defect clustering.(2)The PESI model of spam comments recognition is constructed based on platform,enterprises,society and individual features.On the basis of considering individual(comment content,individual reputation,writing style,star rating),enterprises(business reputation,product characteristics,industry characteristics),platform and society relations(relationship between comment text and reviewer,between enterprises and reviewer,between reviewers),the PESI model with multi-dimensional features is proposed to identify spam comments.(3)A spam comment detection algorithm SD_DCOF is proposed based on density and clustering outlier factor.Applying text clustering and outlier detection theory to spam comment recognition,the detection algorithm SD_DCOF is proposed.Compared with the traditional outlier detection algorithm,it reduces decision-making fraud and improves the reliability of clustering and efficiency of outlier point detection without manual input parameters.The experimental results show that the algorithm can identify spam information very well.(4)The LDA model is improved.Combining the product hierarchy and the internal product defect information,the new LDA model CLDA is constructed to identify and obtain the topic information of product defects,and the qualitative and quantitative methods are combined to evaluate the effect of defect clustering to ensure the quality of the information found.The experimental results show that the model is superior to the standard LDA model and can detect more valuable information.