Software risk item identification is an important activity of software risk management and project management. Along with the development of software technologies and integration, the complexity of identifying software risk items is increased and required signification efforts. Modeling techniques are commonly used to construct prediction models to facilitate risk items identification. The basic assumption of most modeling techniques is that the source domain is the same as the target domain. This study proposes an approach to reduce the impact of software process change while identifying software risk items, in which the clustering technique is applied on the data collected from past software projects to retrieve the knowledge of risk items. The information can be used to construct suitable prediction models for current software project to identify risk items. The advantage of the proposed approach is that the software risk models can be built at early stage of software project to facilitate the software risk mitigation planning.
The 22nd IASTED International Conference on Modelling and Simulation (MS 2011), 加拿大, 2011.7.4