Research On Test Case Reuse Strategy Based On Case-Based Reasoning

With the rapid development of software industry, the size of software systems is expanding ceaselessly, which causes software systems to become more and more complex and makes accidents caused by software feedbacks occur frequently. Therefore, software test, as an important part of software quality assurance, attracts more and more people’s attention. As we know, the choice of test cases, the guide of test execution, is critical to software test. However, because of the differences in experiences and skills, testers sometimes select test cases blindly and the results of test are usually uncertain, which causes the failure of test eventually.Software reuse as an important technique to improve software development efficiency and software quality offers a solution to reduce the test uncertainties. Academia and some industry experts have put the concept of reuse into the practice of software requirements engineering and design engineering, and have achieved remarkable results. Applying this method to software test and making full use of the historical test cases which are accumulated in the past days to design new test cases not only can make up the deficiency of testers’experiences, but also can reduce those redundant works in designing similar test cases, which improves the test efficiency and makes test experiences be passed on. Therefore, the sharing and reuse of test cases has become a research focus of the field of software testing gradually.Test case reuse is to apply historical test cases to new test requirements in varying degrees. In this paper, we study test case reuse and some major test case reuse models first, and then we decide to choose the Case-Based Reasoning(CBR) model. However, because of the low retrieval efficiency, test case reuse status can not achieve the expected degree.As for this problem, this paper presents a classified test case reuse model based on CBR. This model uses a facet stratified method to represent test cases’characteristic properties. Then, we use nearest neighbor algorithm to calculate the similarities of target cases and the historical cases. This model supports retrieval level of single test case, test suit and test project, which allows testers to choose different retrieval sizes in accordance with the various requirements and improves the retrieval efficiency eventually.The significance of this study is having achieved the classified reuse of test cases. Practices have proved that this method does improve the test cases reuse degree, and thus shortens the software testing cycle, reduces the cost of software testing, makes up the deficiency of testers’ experiences to some extent and has practical value.