Using error modeling to improve and control software quality: An empirical investigation

• Main Author: Walia, Gursimran Singh
• Format: Dissertation
• Language: English
• Published: ProQuest Dissertations & Theses 01-01-2009
• Subjects: Computer science
• ISBN: 110908871X; 9781109088717

Software quality, or the lack thereof, is a well-known problem faced by software engineers. To address the problem of poor software quality, many approaches have been developed and evaluated through controlled experiments and case studies in both laboratory and real settings. Considerable effort has been devoted to identify methods to find and repair early life cycle faults i.e., actual mistakes recorded in a requirement or design document. One empirically proven fault-reduction approach is to use fault classification taxonomies to help developers identify different types of important faults. However, even when faithfully applying various empirically-validated techniques that allow developers to focus on faults, they do not help developers to find all types of problems. Furthermore, these techniques can only help detect the presence of faults and not their absence or provide insight into how many faults still remain. To augment the existing methods, and further improve software quality, my research developed and validated new approaches to fill in some of the gaps. This dissertation involves developing and validating effective methods and tools for improving and measuring the quality of software artifacts. A major focus of this dissertation is to understand the thought process of developers so that software quality can be improved. Another focus of my research is to support software defect estimation post-inspection to manage software quality. This dissertation exploits the knowledge about software development errors (i.e., source of the faults) and using the error information to develop techniques that will help developers find and eliminate errors early in the software lifecycle process. This dissertation research is multidisciplinary as it uses approaches that have been applied successfully in other domains and adapts them for the task of improving and managing software quality. This dissertation research added information from cognitive psychology research about human errors to extend taxonomies of software development errors, and also use the Capture-Recapture method (used by biologists and wildlife researchers) to support defect size estimates of software artifacts to manage the software inspection process.