一种基于领域知识的非功能需求建模辅助方法

A Non-Functional Requirements Modeling Aided Method Based on Domain Knowledge

随着信息化的普及,人们对软件系统质量要求越来越高,由于体现涉众对软件质量要求的非功能需求具有抽象、主观、不确定等特性,纯粹依赖人工方式对非功能需求建模是一项十分困难的任务.利用领域工程具有的内聚性和稳定性,提出一种基于领域知识的非功能需求建模辅助方法,通过挖掘隐含领域知识,层次化地组织领域知识,能够弥补个人知识的不足,更准确、完整地建模非功能需求.该方法包含两个部分:(1)基于不确定性理论的领域知识获取及分析模块;(2)基于NFR(Non-Functional Requirements)框架的非功能需求辅助建模模块.通过建立非功能属性分解树、设计基于三角模糊数的评价语言值量化机制和传递依赖关系计算方法,实现了对领域知识的准备;建立了非功能属性及非功能需求间依赖关系的形式化定义和推理规则,提出了基于RE-Tools的NFR扩展,实现了对领域知识的应用.案例研究表明,文中方法具有良好的可操作性,能够辅助建模需求关联,减轻需求工程师的工作负荷;同时,交叉实验结果表明,该方法可提升传统NFR框架方法的效率和准确度.

With the popularization of information technology, more and more attentions have been paid to the quality of software systems. The demands on the quality of a software system are named as NFRs （Non-Functional Requirements）. Abstractness, subjectivity and uncertainty are the remarkable characteristics of NFRs, which brings huge obstacles to requirements engineers to model NFRs manually. Utilizing the cohesion and stability of domain engineering, a NFRs mod- eling aided method based on domain knowledge is proposed. It mainly contains two parts： （1） Uncertainty theory based domain knowledge acquisition and analysis module; and （2） the NFRs modeling aided module based on NFR framework. In the first part, the NFADT （Non- Functional Attributes Decomposition Tree） is constructed to guide the classification of NFRs; the fuzzy set theory is adopted to perform the transformation between qualitative evaluation and quantitative evaluation; and a transitive dependency calculation algorithm is proposed to discover the implicit relationships between NFAs （Non-Functional Attributes）. In the second part, formal definitions and rationale rules of NFAs and NFRs are proposed, and the implementation based on RE-Tools is also presented. Case study shows that the above method is feasible and rational. It is suitable to aid tl^e modeling of dependency relationships among NFRs. Through the cross- validation, it is shown that the method can improve the efficiency and accuracy of traditional NFR Framework. At the end of the paper, the future research directions have been discussed.