XOMO: Processing models within the RST test bed
Tim Menzies · 2006
Abstract Previously, we have defined an iterative data min-ing method for learning better software product and process. That method was prototyped on part of the space shuttle abort software. Here, the same method is applied to artifacts in the Reliable Software Test bench under development at the AMES Research Center. As before, our methods found ways to significantly re-duce risks associated with a project. In particular, the mean estimated defectsKLOC were reduced from 4 to 0.5. Further, the certainty in this estimate was vastly improved since the stan-dard deviation on the estimated defectsKLOC was reduced by two orders of magnitude from 3.4 to 0.05. Better yet, our methods supported a fine-grained analysis of the merits of automated analysis vs execution-based test-ing. When such execution-based testing is impractical, we