Improving and controlling inspections of UML design documents
Shaheen Banu. Boodoo · eScholarship@McGill (McGill) · 2000
Recent evidence indicates that the UML (Unified Modeling Language) is the most preferred and widely used modeling technique for object oriented analysis and design. With UML becoming so popular, there is a need to have good quality assurance techniques for projects using it. Our focus in this study is on the inspections of UML design documents. The basic premise of software inspections is that they detect and remove defects before they propagate to subsequent development phases, where their detection and correction costs escalate. However, the cost performance of inspections can vary considerably, making it important to optimize inspections. Two approaches for optimizing inspections are: (a) controlling the inspection team size, and (b) discovering anomalous inspections that underperformed. This thesis first presents an empirical evaluation of the optimal team size for UML design inspections. Secondly, it presents an empirical evaluation of capture-recapture models for team sizes of 2 to 15 inspectors. Capture-recapture models can estimate the number of remaining defects in an inspected artifact, which can then be used to decide whether the artifact should be reinspected. Our results show that there is no single optimal team size. It in fact depends on various conditions such as the cost of defect detection late in the process, meeting duration and others. This thesis quantifies our findings. And we also found that no capture-recapture models were appropriate for making the reinspection decision as compared to the default decision of passing the document.