Studying Of User Requirements Based On Online Product Community

As the bridge linking enterprises and users,online product community(OPC)has brought together numerous users' complaints or negative reviews about the products.These reviews carry overall,accurate,and reliable evaluation toward products,and can directly reflect the actual user requirements,thus revealing the places where the product needs to be improved,which is of great reference value for the product improvement carried out by enterprises.Therefore,mining and analyzing user requirements in OPC have vitally practical significance.Based on the theory of text mining and machine learning,this paper takes users' reviews toward smartphone in Xiaomi community(XC)as the object and makes a systematic research on the classification of user requirements and the automatic identification of fine-grained user requirements.The main work of this study is as following three aspects:Firstly,a method was proposed to construct the Dynamic Kano Model(DKM)for user requirements.The foundation of DKM was analyzed first;then the index system of user review analysis,which includes the depth and influence of reviews,was established to build the dynamic classification of user requirements in online reviews.On this basis,the overall construction process and key technology of the DKM were studied,and the connotation of its components-Must-be,One-dimensional and Indifferent requirements were analyzed.Finally,the application research was carried out based on the users' feedbacks toward Xiaomi4(X4)in XC.The results of experiment show that the DKM can effectively obtain all kinds of requirements mentioned above.Secondly,a method for automatic identification of fine-grained user requirements was proposed.Firstly,the relationship between the length of the review and the accuracy of prediction for product problem was studied.Then,a method for splitting review sentence was proposed.And the Word2 vec tool was applied to build model which can discover fine-grained user requirements automatically.Finally,the users' reviews of X4 were taken as the object of study,and the auto-labeled product problems were set as the training set of support vector machine(SVM)and then testified whether the related methods were validated.The results of experiment show that the proposed methods can achieve higher recall,precision and F1 values.Thirdly,Application Analysis.Utilizing SVM model,the product problems of X3,X4 and X5 were automatically extracted and identified respectively,and then the change of users' satisfaction in the process of product iteration was studied and the results were analyzed and explained.