Assisted requirements selection by clustering

Abstract

Requirements selection is a decision-making process that enables project managers to focus on the deliverables that add most value to the project outcome. This task is performed to define which features or requirements will be developed in the next release. It is a complex multi-criteria decision process that has been focused by many research works, because a balance between business profits and investment is needed. The spectrum of prioritization techniques spans from simple and qualitative to elaborated analytic prioritization approaches that fall into the category of optimization algorithms. This work studies the combination of the qualitative MoSCoW method and cluster analysis for requirements selection. The feasibility of our methodology has been tested on three case studies (with 20, 50 and 100 requirements). In each of them, the requirements have been clustered, and then the clustering configurations found have been evaluated using internal validation measures for the compactness, connectivity and separability of the clusters. The experimental results show the validity of clustering strategies for the identification of the core set of requirements for the software product, being the number of categories proposed by MoSCoW a good starting point in requirements prioritization and negotiation.