Research On Mobile Online Reviews Based On Text Mining

In the last few years,Along with electronic commerce rapid development as well as the continuous expansion of network transaction scale,consumption behaviors and patterns of consumers have undergone great changes.Considering the convenience,many consumers choose to buy products through online platforms,and then post relevant reviews based on the purchase and usage experience.Further to exploring the deep information of the large amount of relevant reviews producing on Network platforms(including the initial comment information and the additional comment information).The additional comment information may contain more valuable information,we consider the initial and additional comment information,consumers can further understand the performance of the product and the relevant enterprises can obtain important theoretical support providing the improvement direction to improving product quality,optimizing service attitude,enhancing the brand competitiveness.Mobile phones,necessary tools for people to communicating in daily work and life,its performance,appearance and after-sales by consumers service are getting more and more attention.Before buying a mobile phone,people often refer to the products review information on various online platforms.Therefore,digging more available knowledge from mobile phone review information will have an important influence on people's decision-making behavior.Based on the above analysis,the main work of this thesis is as follows :(1)taking the mi 11 mobile phone saled in the self-operated flagship store of jd as an example,the web crawler was used to obtain the comment information of mi 11 mobile phone,including 11838 initial online comment data and 586 additional online reviews;(2)The python software was used to preprocess the raw data.Basing on the emotion dictionary and machine learning classification method,the initial comment text and the additional review text are divided into favorable and bad reviews,respectively.The experimental results showed that the Stacking integrated learning classification algorithm was the optimal one among the three text emotion classification methods with F1,accuracy and recall rate of 0.98.(3)Conducting visualization modeling analysis and LDA themes to the favorable and unfavorable reviews of the initial comment text and the additional review text,respectively,the results indicated that the users of millet had positive attitude to the appearance,pricing,function and the delivery speed of mi 11 mobile phone.But,towarding the after-sales service,the performance of the mobile phone(heat dissipation,life,smooth),the users held negative opinions,among them,the analysis of additional comment text data plays a supplementary role in the analysis of the initial comment text data,helping us to find more advantages and disadvantages of mi 11 phones.Therefore,the conclusion of this research was that analysis the deep information of the large amount of relevant reviews can provide optimization directions to improving the product quality and after-sales service level of mi 11 mobile phones.