On Measurement and Understanding of Software Development Processes

• Activity Number: 169
• Type: Contributed
• Date/Time: Monday, August 4, 2003 : 2:00 PM to 3:50 PM
• Sponsor: Section on Quality & Productivity
• Author(s): Ping Zhang*+
• Companies: Avaya Labs Research
• Keywords: jackknife ; maximum likelihood ; nonhomogeneous Poisson process ; prediction ; software changes ; software engineering

Abstract

Software systems are entities that change constantly throughout their lifetime. Understanding the relationship between different types of changes and the effect of these changes on project outcome is the key problem in software engineering. Primary software change activities involve adding new features and fixing defects. Our fundamental premise is that defects discovered and fixed during development are caused by implementation of new features. Conceptually, this allows us to decompose a software development project into many tiny sub-cycles, one for each new feature. We show that statistical inference under the proposed model is equivalent to the inference problem under a general mixture model with truncated data. We apply the model and the corresponding inference methodology to several historical projects and use it to predict the outcome of a recent project. Our model presents a novel unified framework to investigate and predict effort, schedule, and defects of a software project. The results of applying the model confirm a fundamental relationship between the new feature and defect repair changes and demonstrate its predictive properties.