Analysis Of The Influence Of Testing Coverage On Testing Effectiveness

Software testing is the process of running software in order to find errors in the software.The key problem in software testing is to select sufficient and high-quality test suites to find as much errors as possible in the software.In the research of software testing,code coverage is usually used to measure the sufficiency of test suites.The code coverage metrics include statement coverage,branch coverage,condition coverage,and path coverage.A sufficient test suites should detect more errors in the software,and improving the sufficiency of the test suites is helpful to improve the effectiveness of the test suites.The researchers proposed mutation test,by mutating the original program,evaluating the kill ratio of the test case to the variant,and using the mutation score as the validity of the test case suites Measurement criteria.However,in recent years,the relationship between the test coverage and effectiveness has attracted researchers' attention.This paper researched the relationship between test coverage and test effectiveness.the research group conducted empirical research before,but the main problems are as follows.The first point only analyzes the commons-lang project under the Defects4 J experiment suites,and the second The point is only to analyze the test cases of JUnit 4.In view of the above problems,this paper conducts experiments on all projects under the Defects4 J test suites,researches on test cases based on JUnit 3 and JUnit 4,and proposes a test case selection technique based on balanced coverage in the deep neural network testing.This paper's main research work are as follows:(1)On the six projects of the Defects4 J experiment suites,methods and assertions are used as the smallest unit of test cases to conduct an empirical study on the relationship between test coverage and test effectiveness.By analyzing the experimental results,it is found that as the size of the test suites increases,test coverage and test effectiveness also increase,and test coverage and test effectiveness are highly correlated;but the relationship between the two The influence of the mixed variable of the test suites size,when the mixed effect model based on linear regression is used to remove the influence of the test suites size,there is no obvious correlation between test coverage and test effectiveness.(2)In the testing of deep neural networks,there is a problem that the test sufficiency criterion based on neuron coverage reaches the peak of coverage prematurely,which leads to neuron coverage to guide the test case selection and lose the guidance function.The idea of this paper proposes a deep neural network test case selection algorithm based on neuron coverage.Experimental results on the MNIST dataset show that the algorithm improves the error detection efficiency of deep neural networks.