Information Acquisition for Service Contract Quotations by Repair Shops
Zoom Research Seminar / 5th Floor EE Lecture 2
Past event — 25 April 2023
12:00–13:00
English
Spoken language
Dr. Simon Voorberg
Assistant Professor of Operations and Information Management
Neoma Business School
Abstract
In this talk, we will introduce a model for optimal dynamic information acquisition for profit maximization (DIA model) and show how effective it can be to use this model in an industrial environment to do service contract quotations for repair shops. The process of information acquisition is a strong example of a decision process where information is crucial to make the right decision. The motivation for this comes from a collaboration with Fokker Services where we realized that a very complex policy can be more confusing for decision makers, making it harder for them to accept and adopt this new policy. If we can find more easily comprehensible policies that remain effective, we could sacrifice some profit while having the full support of the executing human decision makers. Based on the DIA model we introduce a refinement of this model that focuses on quotation optimization (QO model). Using the QO model, we introduce a case study example found at Fokker services. Next to finding the optimal policy for this case study, we try to simplify the process in terms of computation time but more importantly in terms of comprehensibility. We show that fixing the order of the attributes has a small effect on the reward of the process, while completely fixing the set of attributes that is revealed has a much larger effect. Furthermore, we show the efficient frontier for these different heuristics compared to the optimal policy when constraining the average invested time.
Bio
Simon Voorberg is an Assistant Professor of Operations and Information Management at Neoma Business School in France. He has a BSc in Industrial Engineering and a MSc in Applied Mathematics. Simon’s research includes optimization of business processes and maintenance optimization in a network setting. Particular methods that are used by Simon are Markov Decision Processes and (deep) Reinforcement Learning. His research has appeared in Decision Support Systems journal and European Journal of Operational Research. His PhD research was in cooperation with international companies such as Philips, Dutch Railways and Fokker Services and focused on real-time decision support for semi-structured data-driven processes.