Research On Data-driven Methods For UI Design And Implementation Of Mobile Applications

In the era of informatization,mobile applications(apps)have permeated into all aspects of people’s life and work.In mobile apps,the user interface(UI)serves as the bridge connecting the application and the user,directly impacting the user’s interaction experience.A well-designed UI interface can enhance user satisfaction and loyalty.A clear and concise layout,easily understandable navigation structure,and pleasing visual effects can assist users in completing tasks conveniently while fostering trust and comfort.The development of mobile application UI involves two crucial tasks: UI design and UI implementation.In completing these two tasks,developers face the following challenges.During UI design,developers need to enhance and optimize the product UI continuously.To that end,they must consistently gather user feedback on the UI and adjust and optimize the UI based on user suggestions.Therefore,how to efficiently obtain user feedback on UI to improve the UI design of application products is the first challenge faced by developers.Additionally,while pursuing UI innovation and uniqueness,developers need to ensure that the function layout aligns with users’ habits.Overly personalized or unique functional layout designs can pose difficulties for users,leading to confusion and resistance.Hence,analyzing whether the function layout of the UI conforms to design conventions is the second challenge for developers.During UI implementation,an important task faced by developers is to write code to implement animations in the UI.UI animation is the visual change that is intentionally constructed within the UI that can enhance its aesthetic appeal while also improve product usability.In the process of implementing UI animations,the majority of programming tasks can be accomplished using animation APIs,thereby avoiding redundant reinvention of the wheel.However,with numerous animation-related APIs available,selecting the appropriate one for the current animation task is an important challenge faced by developers.To address the above challenges,this paper proposes data-driven methods based on natural language processing and deep learning technologies.These methods aim to mine reusable knowledge about UI design and implementation from massive amounts of user reviews and APK files in app stores.They assist app developers in gathering user feedback on UI,evaluating the functional layout of product UI,and utilizing animation API resources.To address these challenges,this paper proposes data-driven methods for Android mobile app UI design and implementation based on natural language processing,data mining,reverse engineering,and deep learning technologies.The primary research content of this paper includes:(1)UI suggestion mining method for user reviewsUI design is a process that requires continuous improvement and optimization.Developers need to constantly collect and analyze user feedback,integrating their suggestions to adjust and enhance the UI.To assist developers in efficiently obtaining user suggestions regarding UI,this paper introduces UISMiner(UI Suggestion Miner),a method for automatically mining UI design suggestions from user reviews and locating these suggestions in the UI.The implementation of UISMiner relies on natural language processing and machine learning technologies,primarily comprising three steps: 1)Training a classifier to identify reviews relevant to UI suggestions;2)Defining rules to extract UI suggestions from the identified comments;3)Associating UI suggestions with the corresponding parts of the UI based on semantic similarity analysis.Experimental results demonstrate that UISMiner can effectively obtain UI-related reviews and extract relevant information from the obtained comments.Human evaluation experiments confirm the usefulness of UISMiner.(2)UI function layout evaluating method based on product design experienceThe functional layout of UI adheres to numerous unofficial yet widely accepted design conventions,including the positioning of functions and their relationships.Failure to adhere to these design conventions in product UI may lead to user confusion during using application.To aid developers in analyzing whether a product’s UI has risks of violating design conventions in terms of functional layout,this paper proposes a functional layout analysis method called Ui Analyzer.This method compares the functional layout of UI with that of similar UIs to evaluate the functional layout of the analyzed UI.Ui Analyzer first generates semantic wireframes for the analyzed UI and similar UIs based on their functional layouts.Subsequently,it utilizes convolutional neural networks to extract visual features from the semantic wireframes and applies anomaly detection algorithms to assess whether the visual features of the analyzed UI exhibit anomalies.Experimental results demonstrate that Ui Analyzer can effectively evaluates whether the functional layout of an application’s UI has risks of violating design conventions.(3)Recommendation method of API list for Android animation implementationIn the process of UI animation implementation,it is difficult for developers to identify the appropriate API to implement the current animation from a large number of APIs.The app market consists of millions of apps that can provide valuable data resources for solving this problem.By summarizing API usage of similar animations in apps,we can obtain reusable knowledge for recommendation efforts.This paper proposes the Animation2 API to mine API knowledge from existing apps and recommend API for UI animation.Unlike existing text-based API recommendations,the Animation2 API uses GIF/video format UI animations as query input.First,by analyzing a large number of apps,we build a database containing UI animations and API mapping relationships.Then,a UI animation feature extractor is constructed to obtain the spatio-temporal feature vector of UI animation.By comparing the spatiotemporal feature vectors between UI animations,the animations similar to query animations are identified from the database.Finally,it summarizes the APIs used by similar animations and recommends a list of API candidates for developers.The experimental results show that Animation2 API can more accurately implement recommendation candidate APIs for animations than text-based API recommendation methods.User experiments have shown that the Animation2 API can improve the efficiency of animation development for developers.(4)Real-time recommendation method for animation APIs based on multimodal informationThe implementation of animations often involves multiple APIs.In order to provide real-time assistance to developers during the programming process,this paper proposes U-A2 A,a recommendation model that recommends animation APIs based on multimodal information from the programming context and animation tasks.This model utilizes 3D-CNN and GRU to extract information from the animation task and API context,respectively,and employs a linear model to predict the next API to be used.By leveraging UI exploration and code analysis techniques to acquire the association between animations and API sequences,this paper trains the U-A2 A model based on these associations.Experimental results demonstrate that U-A2 A can effectively recommend the next API to be used based on API context and animation tasks.In summary,this paper focuses on the UI design and implementation process for mobile applications,employing technologies of natural language processing,data mining,reverse engineering,and deep learning.The research revolves around three key issues: UI feedback acquisition,UI function layout detection,and UI animation API recommendation.The aim is to provide assistance to app developers.