Abstract
AbstractIn the last decade, explainability has been attracting much attention in the machine learning community. However, this research topic extends beyond this field to encompass others such as operations research and combinatorial optimization (CO). This paper addresses this issue in the case of the workforce scheduling and routing problem (WSRP), a CO problem involving human resource allocation and routing decisions. We first introduce a novel mathematical framework that models the process of explaining solutions to the end‐users of a WSRP‐solving system. Then, we present original algorithmic methods to generate explanation texts employing a high‐level vocabulary adapted to such end‐users. Explanations are user‐centered, local, and contrastive. They are triggered by end‐user questions about various topics regarding a solution of a WSRP instance. Both questions and explanations are expressed as texts thanks to templates. Numerical experiments show that the algorithms generating explanation texts have execution times that are mostly compatible with the online use of explanations in an interactive system.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call