Research On Combinatorial Test Set Generation With Parameter Weight

Software testing is an important aspect of software development, which across the entire software development life cycle and is a key to ensure software quality. With the scale and complexity of software system keep on increasing, software testing is becoming a resource-intensive activity, so people have been searching for ways to reduce the cost of software. Past experience and practice show that the combinatorial testing (CT) is a practical and effective software testing approach, it has a high error detection ability due to detect the defects that triggered by the interaction of factors in software under testing (SUT) by a small combinatorial test set.The problem of combinatorial test generation has been a focus of research in the field of CT, which aims to generate a test set as small as possible for a given SUT on the condition of satisfying the combinatorial coverage criteria. In the second chapter of this paper briefly introduces some basic concepts and related definitions of CT in the first, then the several existing algorithms of combinatorial test generation are classified, and analyzes the advantages and disadvantages of these algorithms.In the past, the methods of combinatorial test generation rarely comes to weight values of factors (or parameters) in SUT. However, these factors often need to be endowed with corresponding weight values because of some objective reasons in practical applications. IPO algorithm generates test set by in-parameter-order strategy and has a good scalability, but there are some problems which affects its performance. This paper proposes a method(IPO_PW) based on IPO algorithm to solve the above problem and can generating test set with parameter weight. The third chapter introduces three improvements about some problems that affect the performance of IPO algorithm, including the order of extending of parameters to be extended, the order of extending the existing test set and the choose of the value of parameter to be extended. Because IPO algorithm does not any processing on test set after extending all parameters, there may be some redundant test cases in test set of this time. In order to solve this problem, IPO_PW algorithm will use reduction algorithm processing the initial test set to simplify further. In addition, the entire processing of reduction is also described in detail in the third chapter.The paper based on in-depth research into IPO algorithm, adding the concept of parameter weight. It can generate a test set as small as possible under the promise of satisfying combinatorial coverage criteria and considering parameter weight. Moreover, The fourth chapter gives the overall framework of IPO_PW algorithm, as well as the algorithmic descriptions of some key steps. The experimental results in fifth chapter show that IPO_PW algorithm not only can reduce the size of test set, but also can solve the problem of parameter weight. The former can effectively reduce the cost of software testing. Due to the ultimate test set that is according to the descending order of weights of test cases, the latter can provide an important basis for tester so that they can select test cases. Thus, the practical application ability of IPO_PW algorithm is improved. Finally, analyzing the several problems of existing works in the sixth chapter, and proposing some research directions of future in the filed CT.