Application of reinforcement learning to requirements engineering: requirements tracing

Abstract

We posit that machine learning can be applied to effectively address requirements engineering problems. Specifically, we present a requirements traceability method based on the machine learning technique Reinforcement Learning (RL). The RL method demonstrates a rather targeted generation of candidate links between textual requirements artifacts (high level requirements traced to low level requirements, for example). The technique has been validated using two real-world datasets from two problem domains. Our technique demonstrated statistically significant better results than the Information Retrieval technique.