Research On Multi-attribute Comprehensive Evaluation Of Product Based On Online Reviews

With the development of Internet technology,china's e-commerce industry has realized a leap-forward development,and the type and scale of e-commerce platforms have increased dramatically.The rapid development of e-commerce industry has greatly changed people's way of life.Nowadays,online shopping has become an irreplaceable way of consumption for consumers,and people are keen to choose their favorite products through online shopping.And online reviews,as subjective evaluations of products written by consumers after online shopping,have become an important data asset for E-commerce platforms,these data have provided the huge potential application value for the electronic commerce platform,the consumer,the merchant.Especially for consumers,online reviews are one of the important references for them to make online purchasing decisions.However,how to efficiently integrate and utilize the massive online reviews to form an objective evaluation and understanding of the product,then helps the consumer to make the reasonable purchase decision is worth studying,this thesis has carried on the following work mainly to this question:Firstly,on the basis of literature research on product attribute extraction and weight determination based on online reviews,sentiment analysis based on online reviews and multi-attribute comprehensive evaluation model based on online reviews,etc.,summarizing the achievements and shortcomings of the existing research,and determining the main content of this paper,and giving the overall research framework for the research issues.Secondly,Pkuseg was used for word segmentation and part of speech tagging of online reviews,and the Corpus of online reviews is established.On this basis,product attributes are extracted from the Corpus based on association rules Apriori Algorithm,and the product candidate attribute set is obtained.Then,based on the method of Word Frequency statistics,the word frequency of the attribute words in the product attribute set is counted,and the weight of each attribute is calculated.Tthen,constructing an emotional dictionary,judge the polarity of emotional words,and analyze the modificatory effect of degree words and negative words on emotional words,quantifying consumer sentiment about product attributes in online reviews.On this basis,a new product multi-attribute comprehensive evaluation model based on sentiment analysis combining probabilistic linguistic TOPSIS method is constructed,and the feasibility of the new model is verified.Finally,an example is given based on the multi-attribute comprehensive evaluation model proposed in this paper.Firstly,the emotional tendency of consumers about attributes in online reviews of alternative products is quantified,and getting the emotional tendency of consumers about attributes and its probability distribution,furthermore,the emotional tendency grade and its probability distribution are transformed into the initial probabilistic linguistic evaluation information,which can be used for multi-attribute comprehensive evaluation of products,finally,the comprehensive evaluation value of the candidate products is calculated by the method of probabilistic linguistic TOPSIS,and the ranking of the candidate products is given.The effectiveness and Operability of the product multi-attribute comprehensive evaluation model based on sentiment analysis combining probabilistic linguistic TOPSIS method are further demonstrated through case study.