Multi-Attribute Decision-making Method For Decision Preference Mining And Recommendation In E-commerce Environment

With the rapid development of the Internet and the rapid popularization of e-commerce services,more and more people tend to choose products and services that are suitable for themselves on the mobile Internet.Online shopping has become a development trend and has been fully integrated into everyone’s daily life.However,due to the unique intangibility of online shopping and the synchronization of consumption and production,this will undoubtedly increase the risks of online shopping for consumers.The reviews that consumers post on shopping websites after shopping are comment texts with their own emotions or preferences for the purchased products.For consumers to choose their own goods,in addition to the fixed parameters given by the shopping website,the comment text information published by other consumers in the shopping website is one of the ways to obtain important product information.Therefore,how to dig out potentially valuable information from online review texts to help consumers choose suitable products has become one of the issues that many scholars think about.How to improve the existing hybrid multi-attribute decision-making methods through Chinese text mining technology,so as to further improve the accuracy of evaluation research,is the scientific problem to be solved in this research.After a detailed market survey,the official mall on JD.com is finally selected as the data source for this research.The consumer online comment data of Lenovo Xiaoxin series business laptops in the mall is collected and the Chinese online comment text is taken as the data source.The data is processed by pre-processing operations such as removing duplicate text,word segmentation,removing stop words,and unifying case.Then the processed Chinese text data is input into the LDA subject model,so as to determine the subject parameters of the model using the idea of subject coherence,by means of mining the subjects of the text data through the LDA subject model,as well as obtaining relatively objective evaluation indicators and recalculating the weight of each evaluation indicator.A relatively objective judgment matrix is constructed to replace the judgment matrix which is obtained by the subjective scoring of experts in the original analytic hierarchy process.The Xiaoxin series of business notebooks are evaluated comprehensively by using the constructed consumer online comment evaluation system.The results show that:(1)Among the evaluation indexes of Xiaoxin series of business notebooks,the net weight of products,product price,processor performance and graphics card performance are important indexes that affect consumers’ purchase.Processor performance has the greatest impact on consumer choice.(2)Among the four business notebooks of Xiaoxin series,Lenovo Xiaoxin Pro B ranks first with a comprehensive score of 0.308.Its products are very outstanding in terms of processor performance and graphics card performance,and the overall weight is small.It is recommended that consumers buy them.(3)For Lenovo Xiaoxin Pro A,its comprehensive score is 0.110,ranking last.It has no outstanding product advantages,so it is necessary to create product highlights for it in subsequent product optimization.Based on LDA theme model proposed in this thesis,the improved AHP has certain research significance.Compared with the traditional AHP,it is more objective and accurate,and can accurately build a consumer evaluation system,as well as plays an irreplaceable role in helping consumers choose products and product improvement.