Research On Recommendation Issues For Android Mobile Applications

In recent years,mobile applications have penetrated into every aspect of people's live and work.Whether it's simple communication and entertainment,or sensitive activities like bank transfers,online payments,and e-commerce,people tend to do it at their fingertips.In order to meet people's needs,the number of various types of mobile applications has shown explosive growth.The popularity of mobile applications has also brought huge economic benefits.Compared with traditional software,mobile applications pay more attention to users.It can be said that mobile applications are a kind of user-centric software system.As two important roles in the mobile application ecosystem,software developers and application stores need to complete their respective tasks well to ensure the number of users of products and market,so as to maintain and even expand the benefits of products and market.As far as mobile application developers are concerned,they not only need to realize planned functionalities in the product development phase,but also need to expand the functionalities of products during the iteration of mobile applications.In this process,developers face two key questions:which APIs can support the realization of product functionalities when developing and iterating applications,and which functionalities should be added to the products during application iteration.In addition,with millions of products available,an important task for mobile application stores is to recommend suitable applications to users.As the functionalities provided by applications become more and more powerful,mobile application stores should no longer only consider traditional recommendation factors such as functionalities and popularity.The privacy security of mobile applications also need to be taken into account in the recommendation process.In response to these issues,this paper carries out relevant research based on the technologies such as natural language processing,data mining,and deep learning,and proposes the methods to recommend API information as well as functionality information for Android mobile application developers and to help Android mobile application stores recommend appropriate software products for users.It should be pointed out that the methods proposed in this paper are aimed at Android mobile applications,but the idea of solving the problem is also applicable to other types of applications,such as applications in the Apple Store.The main research contents of this paper are as follows:(1)To help mobile application developers obtain suitable APIs when they realize product functionalities,thereby improving the efficiency of product development and product iteration,and seizing market users as soon as possible,this paper summarizes the API usage experience of existing products in application stores to gain reusable knowledge,and recommends suitable APIs for developers from the level of product functionalities based on the knowledge.Firstly,this paper takes UI(User Interface)components as the bridge to establish mapping relationships between functionalities and APIs.Then,this paper summarizes the functionalities obtained from the UI of mobile applications to establish a functionality framework,and constructs API knowledge for the nodes in the framework based on the mapping relationships between functionalities and APIs.Finally,this paper identifies the corresponding nodes in the framework according to the keywords and expression forms of queries,and displays the API knowledge of these nodes to developers in the form of recommendation lists to help them use the recommendation information effectively.(2)To help mobile application developers expand the functionalities of their products,this paper uses UI pages as data units of information mining to establish the relationships between functionalities,and captures the key functionalities that mobile applications lack compared with similar products based on these relationships,thereby recommending these functionalities to software developers.Firstly,this paper uses UI testing tool to collect the UI pages for mobile applications and gives the method to gain the functionalities in these pages.Then,by comparing the functionality information in the UI pages,this paper identifies the products that have similar functionalities to analyzed applications.Finally,this paper establishes the relationships between functionalities by analyzing UI pages of analyzed applications and their similar products,and recommends suitable functionalities(the missing key functionalities)for UI pages of analyzed applications based on obtained relationships.(3)To help mobile application developers expand the functionalities of their products,this paper also mines the reviews to gain users' functionality requirements and recommends them to developers.Firstly,this paper summarizes the relevant classification factors and trains a classifier based on these factors to classify reviews into reviews containing user requirements and other reviews.Then,this paper defines two types of extraction rules(keyword-based linguistic rules and grammar rules)to extract the functionality information from the reviews containing user requirements.Finally,this paper uses the user attention as reference factor to evaluate the recommendation value of functionalities obtained from reviews and identifies the locations where these functionalities are suitable to be added based on the development experience of existing products,so as to help developers to better use the recommendation information.(4)To help mobile application stores recommend suitable products for users,this paper proposes a personalized mobile application recommendation method considering privacy security and product functionalities.Firstly,this paper mines the functionalities in description texts of mobile applications and summarizes them into aspects that can be used to describe higher-grained functionalities.Secondly,this paper associates user reviews with the aspects obtained and evaluates the completion quality of mobile applications in each aspect by mining user opinions in associated reviews.Thirdly,this paper clusters mobile applications based on their completion quality in various aspects and conducts the security analysis by comparing the sensitive permission usage of products in the same clusters.Finally,this paper infers users' preference on different functionality aspects based on the product download history,and combines the obtained results with users' emphasis on privacy security,so as to complete the product recommendation work from two dimensions: security dimension and functionality dimension.In summary,this paper focuses on three issues of mobile applications: API usage,functionality expansion,and product recommendation.It conducts research based on natural language processing,data mining,and deep learning techniques,and gives the methods that recommend the information(APIs and functionalities)and products,aims at provide help for developers and mobile application stores.