Research On Prioritization And Selection Of Regression Testing Cases Based On Historical Information

In recent years,the United States has imposed sanctions on various sectors of our country,limiting the supply of some parts to our companies.In this case,enterprises seek alternative products from other domestic and foreign markets in order to ensure the follow-up production of products.To ensure that the modified hardware continues to perform the functions of the original hardware without introducing new defects,the hardware and its associated systems need to be tested.Regression testing is an effective method to ensure system reliability,but the testing cost is expensive.Therefore,it is necessary to use effective techniques to reduce the cost of regression testing.Test case prioritization and selection are techniques that can effectively reduce test time consumption and improve test efficiency.Therefore,this thesis studies the prioritization and selection method of use cases suitable for regression testing of hardware change scenarios,which has important theoretical significance and practical value.The main work of this thesis includes two aspects:(1)Aiming at the problems of software source code information not being available,large number of test cases and multiple test versions in the black box regression test scenario,this thesis proposes a regression test case priority ranking method based on historical information.Firstly,from the huge historical use-case execution information,the use-case execution information,execution results and execution frequency of hardware components are analyzed,and component characteristics are analyzed to determine the key factors affecting the relationship between use-case and component.Then,the component difference spectrum is constructed,and the correlation degree between the use case and the component is calculated according to the spectrum difference.Then,the correlation between use cases and components is improved by calculating Hill diversity index.Finally,the correlation degree and diversity index are integrated to sort the test cases.The proposed method is applied to multiple test sets and compared with other classical test case ordering methods.Experimental results show that the proposed method can effectively give the priority of test cases,improve the error detection performance of test cases set,and improve the test efficiency.(2)In view of the problems of poor traceability between test cases and components in the case of multiple changes or unknown changes in the hardware system,as well as the difficulty of selecting test cases by non-adaptive methods,this thesis proposes a regression test case selection method based on reinforcement learning.First,the deep Q network is used as a proxy to assist test case selection.Then,based on the feature that regression testing pays attention to defect detection,an appropriate reward function is designed.On the basis of considering the performance of test cases,the reward for failed test cases is increased,which makes the reward for test cases more specific and helps to assist agents to make better choices.Finally,the deep Q network model is trained by executing historical data sets using successive versions of test cases.By comparing the experimental results,it can be seen that the proposed method can effectively solve the problem of selecting regression test cases,so as to improve the test efficiency.