An Empirical Evaluation of Predicting Runaway Software Projects Using Bayesian Classification

Since software development projects often fall into runaway situations, detecting signs of runaway status in early stage of development has become important. In this paper, we propose a new scheme for the prediction of runaway projects based on an empirical questionnaire. We first design a questionnaire from five viewpoints within the projects: requirements, estimations, planning, team organization, and project management activities. Each of these viewpoints consists of questions in which experience and knowledge of software risks are included. Secondly, we classify projects into “runaway” and “success” using resultant metrics data. We then analyze the relationship between responses to the questionnaire and the runaway status of projects by the Bayesian classification. The experimental result using actual project data shows that 33 out of 40 projects were predicted correctly. As a result, we confirm that the prediction of runaway projects is successful.