Study On A Test-suite Reduction Models With Fault Severity Levels And Its Application

Incorporating fault severity levels, this paper proposes a bi-criteria test-suite reduction model that considers both the coverage degree of test case and the capability of error detection among test cases.Test cases are detained in test suite and organized according to certain coverage degree. The software faults detected by test case were saved into test suite as additional information, and each fault is ranked with a severity level. In regression testing stage, the software faults detected by test cases and their severity levels are acted as the criteria to reduce the test suite, as well as including another parameter - coverage degree. In the process of reducing the original test suite, we introduce a weight factor to allow one criterion to take precedence over the other. By changing the value of the weight factor (0-1), we can obtain reduced test suites with different performance (from the maximum fault-detection capability to the maximum compression ratio). The model can greatly reduce the size of the original test suite and at the same time guarantee the same coverage degree and fault-detection capability. If the performance of either of the two criterions can be sacrificed, the compression ratio could be greater.For example, the model uses a group of binary decision variable to represent if the test case is needed for a requirement; another decision variable is introduced to record the sum of the severity levels that belong to the faults detected by the test case. As long as the fault-detection information is timely updated after a test case is ran, the model can guarantee that the obtained reduced test suite has excellent fault-detection capability, as well as it can cut down the operational cost, satisfy all the requirements and maximize the fault-detection capability.The bi-criteria test-suite reduction model with severity level can significantly increase the efficiency of the testing process, also provide more detailed information to evaluate the capability of detecting fault among the test cases, all these is helped under a unique characteristic of the model, that is the reduced test suite satisfies adequate coverage degree no matter when the algorithm is terminated. Practice proves that the model can not only effectively reduce the original test suite, but ensure the new test suite to hold the maximum error detecting capability. Accordingly, under the premise of guaranteeing software quality, the testing cost is significantly reduced.