The Study On Adaptive Random Testing Based On Filtering Test Set And Its Application In Numerical Programs

In the information sociality,software is increasingly widely involved in human activities as a key component of the information system.And software testing has been one of the significant processes of the software lifecycle and plays an important role in ensuring software quality.Among various testing methods,Random Testing(RT)has been widely implemented in many fields due to its simple and easy deployment.However,random testing only relies on the random algorithm without using any known information,it presents poor fault detection capability.To improve the effectiveness of RT,TY.Chen et al.took the characteristics of failure input continuity,and proposed Adaptive Random Testing(ART).ART has been proved that it achieves greater effectiveness than that of RT by evenly distributing test cases across input domain.Among numerous ART algorithms,the Fixed-Size-Candidate-Set Adaptive Random Testing(FSCS-ART)has been regarded as a popular method with high testing effectiveness.However,FSCS-ART requires to calculate a large scale of test case similarity when selecting the optimal candidate from the candidate set,which has brought heavy computational overhead to a certain extent.This disadvantage limits its practical application.To enhance the efficiency of FSCS-ART,this paper focuses on the test case selection process of FSCS-ART,and two algorithms are designed based on the filtering mechanism to reduce the number of candidate test cases and the executed test cases involved in calculating test case similarity.Finally,we have designed and developed a prototype system to analyze and evaluate the effectiveness and the efficiency of our proposed methods.The experimental results show that the proposed method have improve the testing efficiency of FSCS-ART while maintaining the great effectiveness.The main tasks of this paper are scheduled as follows:1?To reduce the heavy computational overhead caused by randomly generated candidate test case,we have proposed the FSCS by Candidate Test Set Reduction(FSCS-CTSR).The key idea of FSCS-CTSR is to store and carry forward some candidates that with low similarity into next test case selection,and other candidate test cases are filtered off.In this process,the scale of candidate set is unchanged.Thus,in the next round of test case selection,compared to the randomly generated candidate test case,the similarities of stored candidate test cases have been calculated already,so that the time overhead can be reduced.According to the results of simulations and real-life program testing,it shows that FSCS-CTSR reduces nearly half of time overhead without the loss of failure detection capability of FSCS-ART.2?For the executed test case set,we have designed two distance constraints criteria including: 1)part of the verified distance;2)approximate distance.And we have proposed the FSCS by Executed Test Set Reduction(FSCS-ETSR).In the process of finding the nearest neighbor for a candidate(test case similarity calculating),FSCS-ETSR firstly calculates the constraints criteria for an executed test case,and then compares the criteria with current minimum distance to determine whether this executed need to be filtered.Thus,the computational overhead can be reduced.The experimental results show that under low-dimensional input domain conditions,FSCS-ETSR with part of the verified distance achieves greater efficiency while maintaining the same fault detection capability of FSCS-ART fault;under high-dimensional input domain conditions,FSCS-ETSR with approximate distance present better testing performance than that of FSCS-ART.3?We have designed and implemented efficient adaptive random test prototype systems based on filtering mechanism.The prototype system consists of an algorithm setting module,a simulation test module,a real program test module,and a statistical analysis module to achieve a high degree of automation.This research is based on the realized prototype system to verify the effectiveness and the efficiency of the proposed algorithms in numerical program testing.