Software Project Level Estimation Model Framework based on Bayesian Belief Networks

Software estimation models should support managerial decision making in software projects. We experience that most of current models do not achieve this goal to the extend managers are looking for. This paper presents a software project level estimation model framework based on Bayesian belief networks. The framework is constructed by using four basic BBN sub-models, component estimation, test effectiveness estimation, residual defect estimation and test estimation sub-models. The integration of these submodels achieves an estimation model suitable for project levels. With this project level estimation model, the estimation and analysis of quality, effort, schedule and scope can be carried out at both project level and specific project phase level. We show how this approach is used in a sample project, allowing project manager to implement an initial estimation, trade-off quality, effort, schedule and scope, and refine the estimation in the later phase of the project