The complex and emergent behavior of software systems makes information and data key components of this unpredictable environment. The use of a data-driven approach to identify and to accurately predict the sources of software project delays, cost overruns, failures, or successes may prove a significant contribution to the fields of systems engineering, software development and project management. Software project failures are pervasive and despite the research, failures still persist. The research presented highlights the systems mindset in addressing failure to introduce a software-specific predictive analytics model that accurately predicts software project outcomes of failure or success and identifies opportunities for incorporation in the federal and commercial space. The research describes how to systematically learn from historical software project failures to train a confidence level model to make project failure predictions. The research demonstrates how to train a generalized model with a range of failure factors combined to efficiently and accurately predict software project failures. The results of the model would be used during acquisition, prior to project initiation, and throughout the software development lifecycle. It is a decision analysis tool to assist decision makers in making the crucial decisions early in the lifecycle to cancel a project predicted to failure or to identify and implement mitigation strategies to improve project outcome. The use of an evidence-based approach to identify software project failure factors will result in better understanding of these phenomena that will ultimately improve software project success rates and minimize risks in systems engineering efforts.
|In Administrative Set: