Research On User Review Classification And User Contrition Measurement In APP Evolution

With the rapid development of mobile Internet and mobile devices,mobile applications(APP)have spread all over the life scenes and become an inseparable part of people’s daily life.According to statistics,the total number of APPs in China’s local third-party APP stores and Apple App Store is more than two million,and the huge demand of APPs brings infinite opportunities and great challenges to APP developers,mobile application stores have obvious public characteristics,and the function descriptions and update documents of a certain APP are publicly visible,which can be easily imitated by similar APPs,while the market competition is extremely fierce,and there are many APPs with highly similar functions,which are highly substitutable.There are many similar APPs with strong substitutability.Therefore,only by satisfying users’ requirement and providing them with better experiences continuously can the competitiveness of the APP be maintained in the market for a long time.Different from the consumer software in the traditional PC era,the existence of the APP store provides an excellent interactive channel for APP developers and users.A large number of users purchase or obtain the required APP software for free through the APP store,and can easily leave a review or give feedback on the APP.The developers also update the APP in time through the APP store and publish the update log.The evolution history of APP and users’ feedback on APP have automatically become part of the APP store,which provides a new way for developers to acquire the requirements of popular users,and also a pioneering attempt to explore and measure the contribution of the general public’s requirements.However,the large number of user reviews,the low proportion of valid information,and the scattered focus of users make it difficult to analyze systematically,and effective approaches are needed to classify and organize user reviews in order to support the discovery and mining of real user needs from user reviews and the analysis of the contribution of popular user requirements in the evolution of APP.In response to these problems,the main work and innovative contributions of this paper are as follows.1.In view of the characteristics of short user reviews,low proportion of reviews with valid information,unstructured expression,and scattered user concerns,this paper proposes a general user review classification approach,which not only alleviates the defects of the current user review classification pre-processing that is cumbersome and dependent on review corpus,but also can be applied to multilingual user review classification,and the experimental results show that the approach effectively improves the current user review classification,which can provide better support for discovering valid user requirements from mass user reviews.2.This paper innovatively carries out the exploration of the contribution measurement of mass user requirement in the evolution of APP.This paper proposes a time-aware analysis approach,which attempts to mine and discover the contribution of popular user requirements to APP requirements engineering in the development and evolution of APPs from user reviews and APP update logs.The core idea of the approach is to take the correlation between user reviews and their time points and the updated content and time points in the APP update log as the main line,and try to effectively mine and extract the popular user requirements adopted by APP developers from user reviews,and combine the new requirements in the APP update log to get a preliminary ratio of popular user feedback to the total APP requirements.To this end,firstly,the aforementioned user review classification approach is applied to automatically classify user reviews,and on this basis,deep learning algorithms are used to match the updated features in APP update logs with the relevant requirements extracted from user reviews.Then,according to the matching results,dig out the relationship between the new functions in the APP update log and the requirements from user reviews,and obtain the feature feedback pattern between developers and users in the development and evolution of the APP.Finally,based on these patterns,we analyze the degree of adoption of user requirements by developers in the process of APP development and evolution,and then explore the influence of users on APP development and evolution.