(An) automated requirements prioritization approach based on stakeholder needs representativeness

Abstract

Recently, the size of software is getting bigger as the software product market rapidly grows. Expeditiously growing software product market involves the large scale software development projects with massive amount of requirements. Having hundreds, sometimes even thousands of individual requirements in a single project is not unusual. On the other hand, in almost every software development project, time to market constraints and resource restrictions make implementing the entire set of requirements not feasible. Therefore, software requirements should be prioritized and be implemented in order of importance. Nevertheless, requirements prioritization is recognized as a difficult and very challenging activity. The limited scalability of requirements prioritization methods, especially in large projects, is repeatedly pointed out in the literatures. The reason of this problem is that the most of the existing methods depend on highly complex and time-consuming human-made manual decisions. To address this scalability challenge, we propose a fully automated requirements prioritization approach, ToMSN (Topic Modeling Stakeholder Needs to prioritizing requirements), which substitutes the manual requirement importance assessment with automatic analysis of stakeholder needs representativeness―how much of stakeholder needs a requirement represents―via topic modeling technique. The approach identifies the requirements that can satisfy the most of the stakeholders when implemented, thus having greater value to business success and return on investment, by analyzing requirements’ stakeholder needs representativeness. Accordingly, the requirements with high stakeholder needs representativeness would be ranked high in the prioritized result and vice versa. To the best of our knowledge, this is the first study to solve the limited scalability problem of requirements prioritization by fully automating the entire prioritization process. A large-scale real software system development project for an approximately 30,000-users was utilized for the performance evaluation and ToMSN approach showed competitive prioritizing accuracy with existing manual approaches while getting no aids of manual decision made by human, therefore ultimately solving scalability issue.