現行開發軟體專案的困境在於軟體專案開發過程中,存在許多的不確定(uncertainty)因素。這些不確定的因素影響軟體專案的品質,同時也影響軟體專案資源的有效應用。 本研究針對這個議題,以貝式網路為基礎,提出一套BBN為基礎的風險管理方法,主要包含事先專案風險評估技術(risk assessment beforehand)、進行中專案資源調整技術(resource adjustment using execution profile)以及軟體度量技術(metrics)。事先專案風險評估技術主要在於軟體專案進行前對於特定的議題,進行what-if的分析與評估。進行中專案資源調整技術針對軟體專案開發過程中的資源進行排序與調配,達到動態調整資源的目的。專案度量技術針對傳統結構式軟體度量與物件導向軟體度量提供一更客觀評估軟體專案的方式。本研究所提出的方法可提昇軟體專案發展效率,並增加軟體專案的成功機率。
The difficulties of software project development is that there exist many uncertain factors during development process. Such uncertain factors may affect the quality of software products, and the effectiveness of the software process. This thesis addressed the issue by proposing dynamic software project management using BBN. Based on modularized Bayesian Belief Network, we proposed a BBN-based project risk management methodology, including risk assessment beforehand, dynamic resource adjustment using execution profile, and software metrics technique. Our risk assessment technique can be applied to answer what-if questions for particular issues before the project starts. Our resource adjustment technique dynamically readjust resource utilization based on resource cost-effectiveness during project execution. Our metrics technique using BBN focuses on traditional software metrics and object-oriented metrics, and provides a more objective assessment for software projects. The proposed techniques enhance the efficiency and success probabilities of software projects.