New paper on managing distributional shifts published in Production and Operations Management
The paper “Addressing distributional shifts in operations management: The case of order fulfillment in customized production” has been published in the flagship journal Production and Operations Management. The paper is co-authored by Julian Senoner, Bernhard Kratzwald, Milan Kuzmanovic, Prof. Netland, and Prof Feuerriegel. The research was conducted in cooperation with the engineering company Aker Solutions in Norway.
This paper focuses on the challenge of optimizing production schedules in manufacturing when customized products are involved—leading to so-called distributional shifts in operational data. To improve production scheduling, the authors propose a data-driven approach based on adversarial learning to address the distributional shifts. The objective of the research is to predict throughput times on production lines given order specifications. Distributional shifts can negatively impact the performance of predictive models when applied to future customer orders with new specifications.
The authors validate their proposed approach using real-world data from a job shop production that supplies large metal components to an oil platform construction yard. Through a series of numerical experiments, the authors compare their adversarial learning approach to common baselines and find that it outperforms them. The paper demonstrates how production managers can enhance their decision-making processes under distributional shifts by adopting this approach.
This paper is among the first to address distributional shifts in operations management. This research received financial support from the Norwegian Research Council through the project COM-FLEX.
We congratulate the authors on the excellent publication. It is available as open access external page here.
