Machine-learning based sensitivity analysis for a sustainable supply chain optimization model

Abstract

The company's operational research allows the identification of improvement opportunities, process critical stages of the process and relevant information that allows a better understanding of the whole process. When analyzing the supply chain, it is possible to find different approaches that create an upturn in the synchronization between the participants, achieving better response times with a superior capacity to face the fluctuations in the demand. The complexity of the supply chains increases when there is a high variability at some level, which is why the economic and geopolitical aspects must be taken into account when planning the internal processes. This presents the challenge of finding a feasible and sustainable solution which guarantees the demand requirements for the supply chain. Also, there's a need to create an effective data analysis methodology where the decision makers can take advantage of the process knowledge, their criterion and experience to focus on the sensitive echelons of the system. The research objective was to implement machine-learning algorithms to analyze the sensitivity of the supply chains with a sustainable scope. This included integral metrics on the optimization process and modeling recommendations for the Sustainable Development Objectives. Additionally, interactive tools were developed to solve an optimization problem based on linear programming and different criteria, including economic, environmental, and financial considerations. The above is framed in a case study of a Colombian textile industry that needed to analyze investment and operational alternatives with a multi-objective scope that guarantees good performance and stability. A 663% improvement in the expected net present value results was achieved. In addition, operational alternatives were found that not only improved the expected outcome of various metrics, but also improved system dynamics and resilience (headroom) with waste reduction over the entire time horizon. In conclusion, a data-based methodology is proposed to analyze the sensitivity of sustainable supply chains, giving the opportunity to identify valuable insights, whose can guide the decision-making process to a better performance and/or less variable results in the different objectives that are studied.