Mining Complex Requirements Specifications to Mitigate Risk via Clustering

Requirements engineering for complex systems in resource-constrained environments produces an intricate set of dependencies. Finding ways to feasibly attain those requirements while adhering to resource limits can be extremely challenging. Here, we describe an approach that helps decisionmakers better explore this complex maze of data. The approach rests upon a combination of existing technologies drawn from the computer science milieu – most notably, heuristic search and clustering, coupled with appropriate visualizations. The context for this work is a requirements engineering method in use at the Jet Propulsion Laboratory, where its primary application area has been planning the development of spacecraft technologies. In this context there is a need to assess the range of risks that threaten to impede requirements attainment, and to plan for their satisfactory (i.e., cost-effective) mitigation. Because of cost and resource constraints, project managers are forced to make difficult choices among interrelated sets of mitigations activities. We show how the use of heuristic search reveals the overall cost-benefit space of requirements attainment, from knowledge of which the project managers can identify feasible solution neighborhoods. Within those neighborhoods clustering is then applied to distill from a plethora of solutions a manageable number of distinct solution categories. Visualizations are used to present this information in ways that assist managers to make their key decisions.