Research On User Demand Mining Based On Feature Extraction And Emotion Analysis

The accurate acquisition of user demand information is an important part in product planning and positioning,and plays a decisive role in the success of products in the market.Traditional methods of obtaining user demand research data such as questionnaires or eye movement experiments have certain disadvantages in terms of difficulty level,quantity,research cost and reliability of research results.The goal of research in this area is still to quickly and easily capture user needs and requires further exploration by scholars and enterprise user researchers.With the development of Web 2.0,the web platform generates a large amount of user-generated content,which is a personal opinion published by users after purchasing or using products/services,reflecting the user’s experience and true feelings.User-generated content has gradually become an important data resource for demand research based on its large amount and convenient access.In order to make up for the shortcomings of traditional user demand research,this paper proposes a user demand mining model based on feature extraction and sentiment analysis.The main research contents include the following aspects:(1)Review usefulness classification model construction.Useful classification of online reviews for user demand based on product perspective.On the basis of literature research,the usefulness classification model indicators are selected,and the effects of classification algorithms such as SVM,KNN,decision tree and random forest are comprehensively compared,and random forest is selected as the basic algorithm of classification model.Based on the training set and test set,the initial model is optimized and the final review usefulness classification model is constructed.(2)Implicit features and implicit sentiment analysis.This paper proposes new implicit features and implicit sentiment analysis methods based on explicit feature sentiment sentences,implicit feature sentences and implicit sentiment sentences,combined with feature lexicon,1V1/1Vn emotion-feature rule base and Doc2 vec statement similarity calculation.The model performs feature-emotional word pair extraction according to the extraction rules.Based on the feature-emotional word pair extraction results,the emotional lexicon and degree adverb polarity values are used for sentiment analysis to obtain the user’s emotional attitude towards product features.(3)User demand mining and classification.Based on the preliminary analysis results of feature extraction and sentiment analysis,combined with the KANO model to establish user demand mining rules.Considering user attention and user satisfaction,the calculation method of user demand priority is proposed to obtain the final user demand list.Different recommendations are made for product design or improvement based on user demand category and priority.(4)Empirical research and analysis.This paper selects Huawei sports bracelet as the research object,and uses reptile technology to obtain users’ comments in the electronic shopping platform and product community as data sets.Based on the two data sets,the characteristics of feature extraction,sentiment analysis,user demand mining and classification,and priority determination are analyzed,and the empirical results are compared to analyze the difference of demand among different user groups.The empirical research validates the feasibility of the user demand mining model proposed in this paper.The results show that:(1)The accuracy and recall rate of the useful classification model based on the product perspective are both 84%,and the model effect is better than other classification models;2)Based on the explicit feature research,this paper increases the extraction and analysis of implicit features and implicit emotions,enriching the results of user demand mining,and the data volume is increased by 11.01%;(3)Different types of user demands have different priorities,and are related to user attention and user satisfaction;(4)Electronic shopping platform and product community users both have higher attention to function parameters,specifications,appearance,battery and APP.But users of e-shopping platforms pay more attention to the workmanship and price of products,and users in the product community pay more attention to the quality of the display and connection problems between the sports bracelet and the mobile phone.Combined with the above experimental analysis results,it is recommended that enterprises comprehensively consider the type and priority of user demand for product design or improvement,and formulate different marketing strategies for different user groups to achieve better marketing results and achieve maximum benefits.