Non-functional Requirements Prioritization: A Systematic Literature Review

Continuous delivery and rapidly changing requirements in agile environments force the developers to put non-functional requirements (NFRs) on halt till maintenance phase. However, neglecting NFRs during prioritization phase may lead to inaccurate estimations for software projects resulting in high maintenance cost and failures. The subjective and uncertain nature of non-functional requirements makes them unfit to be prioritized using conventional prioritization methods. Although the existing literature reports on inadequate consideration given to NFRs prioritization, still no comprehensive systematic effort has been done to report the limitations and evaluation mechanisms of existing NFRs prioritization approaches. Requirements engineering society lacks a broad understanding of NFRs prioritization approaches and the challenges which need to be overcome. Therefore, we aim to investigate (i) the existing NFR prioritization techniques and their validation mechanisms, (ii) the role of Artificial Intelligence (AI) in NFRs prioritization, and (iii) the limitations of existing NFRs prioritization techniques. For this, we reviewed the literature published from 2008 till present and extracted 30 studies. The results reveal twenty-five NFRs prioritization techniques out of which only three are AI based. The major limitations we have come across are that most of the NFRs prioritization techniques are not scalable to large datasets, inter-dependencies between functional requirements (FRs) and NFRs are ignored, and the uncertainties associated with NFRs are not considered at all. However, the literature suggests that AI-based techniques and Fuzzy logic may be used to solve issues such as uncertainties i.e. ambiguities, vagueness, and subjective opinions of stakeholders. This review adds to the existing body of knowledge on NFRs and motivates the practitioners to focus on the NFR prioritization by highlighting the limitations of the existing methods.