An approach to constructing feature models based on requirements clustering

Abstract

Feature models have been widely adopted in software reuse to organize the requirements of a set of similar applications in a software domain/product line. However, in most feature-oriented methods, the construction of feature models heavily depends on the domain analysts' personal understanding, and the work of constructing feature models from the original requirements of sample applications is often tedious and ineffective. This paper proposes a semiautomatic approach to constructing feature models based on requirements clustering, which automates the activities of feature identification, organization and variability modeling to a great extent. The underlying idea of this approach is to analyze the relationships between individual requirements and cluster tight-related requirements into features. With the automatic support of this approach, good quality feature models can be constructed in a more effective way. A case study is also provided to show the feasibility of this approach.