Predicting Changes in User-Driven Requirements Using Conditional Random Fields in Agile Software Development
Ming Qiang Yin, Danli Meng, Dan Zhu, Yibo Wang, Jijiao Jiang · IEEE Transactions on Engineering Management · 2021
Agile development encourages requirements change. The accurate predictions of changes in user requirements could help software evolve in the right direction to increase user satisfaction. Previous research on requirement predictions mostly relies on historical defects or user feedback. In this article, we aim to predict requirement changes based on user–system interactions in agile software development. We focus on the user–system interaction behaviors used to infer user intentions and predict requirement changes to drive the incremental iterations of agile development. Through a prototype system with two incremental iterations, an embedded program in the system captures the user runtime interaction behavior data. We utilize the conditional random fields to explore the user potential intentions and infer the emerging requirements accordingly. The increased accuracy of results in the iterations proves the effectiveness of our approach in predicting user requirement changes.