Theoretical Research Of Combinatorial Testing And Adaptive Random Testing

Software testing is one of the most important methods to assure the quality of software. The modern software system needs to be carefully tested before going into service. If the software is not carefully tested, it is very likely to go wrong and cause substantial financial losses. Therefore, software testing is very necessary. But the problem is that exhaustive testing costs too much and incomplete testing can not guarantee that there is no problem in the system. So, a trade-off of testing costs and effectiveness is necessary, which means we should choose less test cases while keeping the high effectiveness.Combinatorial Testing(CT) and Adaptive Random Testing(ART) are widely used testing algo-rithms. CT thinks the errors are triggered by combinations of factors(or parameters). For the soft-ware with k factors, it's no need to cover all possible k-tuple combinations of parametric values, r-way combination of factors could be satisfied the testing requirement(compared with k, ris a small number). Random Testing(RT) is fast, simple and easy to be implemented, it is often treated as a supplement to other test methods. ART take the advantages of RT, select test case by a certain rule to make sure the test cases are well distributed in the whole solution space, this method enhanced the fault-detecting ability of RT. This paper aims to study the fault-detecting ability of Combinatorial Testing and Adaptive Random Testing in a variety of scenarios.This paper proposes one way to analyses the fault-detecting probability of CT in Boolean spec-ification Testing by minimal failure-causing schema. Through the analysis of the 20 boolean expres-sions extracted from TCAS II system, get 19131 mutants by 10 mutant types. Then all passed test cases and failed test cases are identified for each mutant expression and utilized to get all minimal failure-causing schemas for each mutant. Finally, the fault-detecting probability is calculated for some strengths of Combinatorial Testing. The results show that the CT is very effective in Boolean specification Testing.Then this paper focuses on analysis of fault-detecting ability of ART. The simulation experi-ments are settled that analyses ART in 2,3 and 4 dimensional continuous space with different size of faulty area, two errors are implemented into the algorithm at the same time. The RT is also performed for comparison. The results show that ART is better in 2 dimension than in 3 or 4 dimension. The internal errors have great influence on ART while the modeling error does less influence. In 4 dimen-sion experiment, ART is significantly worse than RT, which means ART is not fit the high dimension scenes.Finally, this paper proposes a strategy to improve the efficiency and effectiveness of ART al-gorithm, Feature Selection. A set of simulation experiments are performed. After that, this strategy is utilized in Boolean specification Testing to verify the feasibility. The results show that feature selection strategy is well performed in complicated scenarios and Boolean specification Testing.