User Behavior Pattern Mining And Reuse Across Similar Android Apps

Nowadays,Android applications have penetrated into all aspects of people's lives,providing people with many conveniences.With the increasing number of Android apps and their increasingly complex functions,software testing of Android apps has become a research focus in the Android field.Understanding the functions of Android apps is the foundation of the Android testing process.It aims to deeply explore the app's business logic,which is especially important in testing.At present,the most popular way to understand an Android app's functions is the user interaction with the app in actual usage scenarios,forming user traces.Based on user traces,user behavior patterns can be mined to understand the app's functions and assist with testing.However,due to the high labor costs and the long test cycle,the testing efficiency of Android apps is seriously affected.Based on the concept that similar Android apps sharing user behavior patterns,our study proposes a user behavior pattern mining and reuse approach across similar Android apps.The approach's core idea is to reproduce the user behavior patterns of a similar known Android app on the Android app under test(AUT),realizing the sharing of similar user behaviors,which can assist the automatic function understanding of AUT and improve the efficiency of subsequent testing.Specifically,our study aims to Android apps,adopting sequence pattern mining technology to mine user behavior patterns from user traces based on their sequence characteristics.On this basis,our study proposes the semantic-based cross-app event matching strategy and the AUT's GUI model-based optimal path selection strategy to realize the across-app user behavior patterns reuse,providing effective support for the AUT's subsequent functional testing.In order to prove the effectiveness of user behavior pattern mining and reuse approach across similar Android apps,a series of experiments have been conducted on 16 Android apps from four categories.The results show that for the total 40 user behavior patterns of experimental subjects,the semantic-based event fuzzy matching strategy can effectively improve the accuracy of event matching.On average,88.5%of user behavior patterns can be completely reused across similar Android apps,and the reused behavior patterns cover 88.0%of the important states of AUT.The above results prove that the proposed user behavior pattern mining and reuse approach can effectively assist the functional understanding of AUT,laying a foundation for its testing.