Multi - Fault Location And Analysis Technology Of Software Based On Data Mining

As the complexity of software continues to expand and the scale of software continues to grow, the structure of the software becomes more and more complex. Reliability of the software becomes increasingly important, the loss caused by software failure occurs numerously. Software fault localization technology has made great stride. There are many fault localization technologies, including test based fault location (TBFL) technology that uses test suit coverage information to locate faults. At this stage, researchers are mostly concerned with single fault position technologies, few studies about multiple faults localization technologies are proposed. This paper proposes a method that uses single fault localization technique to locate multiple faults in parallel mode. Based on previous studies of other researchers, this paper proposes a structure for our method and the paper gives the main component of the structure a detailed introduction, meanwhile imports clustering to the method. The key of multiple fault localization is how to construct a specialized test suit that assists in locating a particular fault and this paper proposes a coverage vector based test suit constructor that cluster the coverage vector to clusters that cause specific fault to fulfill the key task, with the help of clustering, this paper produces specialized test suits to debug multiple fault in parallel simultaneously.Based on the specialized test suits, this paper employs Tarantula, a one-bug-at-time method, to introduce the theory of multiple fault localization. Considering the poor efficiency of Tarantula, this paper combines a cross-tab based fault localization method with parallel fault localization theory to create a new multiple fault localization method and performs experiments on it.At the end of the paper, the paper gives a detailed introduction of the experiments. The paper injects faults into Siemens suits and carries out experiments on the multiple fault versions. With the result of the experiment, the paper compares the cross-tab based multiple fault localization method with Tarantula based multiple method.