AN APPROACH BASED ON MACHINE LEARNING AND DISCRETE EVENT SIMULATION FOR SUPPLY CHAIN OPTIMIZATION - THE CASE OF ON STOCK CHAINS - Research Paper

Aug 4, 2026

Upload by Skyler

Supply chain optimization for on-stock chains faces high complexity driven by diverse, competing performance indicators and by decoupling between procurement and expedition processes. The study aims to mathematically connect on-stock supply chain evaluation parameters to operational action parameters using machine learning, while addressing the lack of labeled data. Discrete event simulation generates scenario outcomes with inherent labels for training and prediction. The paper tests and compares multiple machine learning algorithms to formalize delivery-delay modeling and support optimization decisions.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages20

Reading Time50 min

AccessFree to download

Introduction

On-stock supply chains and the decoupling challenge

Discrete-event simulation to generate labeled data

Bridging simulation and machine learning with a mathematical model

AN APPROACH BASED ON MACHINE LEARNING AND DISCRETE EVENT SIMULATION FOR SUPPLY CHAIN OPTIMIZATION: THE CASE OF ON STOCK CHAINS Zineb Nafi*, Fatima Ezzahra Essaber, Fatine Elharouni, and Rachid Benmoussa Department of Industrial and Logistic Engineering National School of Applied Sciences, Cadi Ayyad University Marrakech, Morocco *Corresponding author’s e-mail: [zineb.nafi@ced.uca.ma](zineb.nafi@ced.uca.ma) The complexity of supply chain problems, more specifically the case of on-stock chains, is due to performance indicators variety, antagonism, and the difficulty of understanding the effects and interactions of different performance drivers with regard to these indicators. As mathematical formalization is essential to optimize the performance of these chains, this paper generally aims to study the contribution of Machine Learning to mathematically link the evaluation parameters of an on-stock supply chain to its action parameters. This work is based on an academic case study that seeks to mathematically formalize the problem of delivery delay in an on-stock supply chain. To this end, several Machine Learning algorithms have been tested and compared. This experience highlighted the impossibility of obtaining a labeled dataset through data collection from the real system. It thus demonstrates the necessity to use a simulation system, in particular, discrete event simulation, to generate this dataset. Keywords: Machine Learning, On-stock Supply Chain, Mathematical Formalization, Simulation, Optimization. (Received on October 5, 2022; Accepted on October 27, 2023) 1. INTRODUCTION Supply chain management is a complex field with various challenges, including cost optimization, quality control, and meeting strict delivery schedules. In the context of on-stock supply chains, a specific procurement strategy is employed where purchased items are stored before being sent to customers. However, this approach introduces a phenomenon known as"decoupling," creating a disconnect between the procurement and expedition processes. While individual procurement actions may not directly affect specific expeditions, their cumulative impact significantly influences overall performance. To address these complexities, we utilize Discrete-Events Simulation (DES), a powerful analytical tool capable of dissecting and understanding each link of the supply chain independently, without requiring predefined connections, instead of Machine Learning (ML), which relies on labeled data and direct links. A labeled dataset consists of data points that are explicitly categorized or "labeled" to indicate their characteristics or outcomes. Moreover, due to the inherent decoupling problem within on-stock supply chains, collecting such labeled datasets becomes challenging. The decoupling creates discontinuities and uncertainties that make it difficult to directly label data. Asa result, applying machine learning techniques, which rely on these labeled datasets for training and prediction, becomes a complex endeavor in this context. To address this challenge, we turned to discrete event simulation (DES) as part of our methodology. DES allows us to model and understand each individual link within the supply chain, offering a clear and accurate representation without the need for pre-labeled data. By simulating various scenarios and events within the supply chain, DES generates data that inherently carries labels associated with each step and outcome. These outcomes serve as the labeled data points we require for training and informing our machine-learning models. To comprehensively address the challenges within the global on-stock supply chain, we recognize the need for a dual approach. DES allows us to precisely represent and understand each link of the supply chain individually, offering a clear and accurate depiction. On the other hand, ML thrives when provided with labeled data sets. By combining the strengths of DES and ML, we can create a

FAQs

Why is optimizing on-stock supply chains difficult?

Optimization is difficult due to multiple, antagonistic performance indicators and the decoupling phenomenon, which disconnects procurement from expedition.

How does the approach obtain the labeled data needed for machine learning?

It uses discrete event simulation to model links and events, producing outcomes that inherently provide labels for each step and result.

What is the main objective of the paper?

To build a mathematical model that links discrete-event simulation with machine learning by bridging evaluation parameters and action parameters, focusing on delivery-delay formalization.

Aug 4, 2026

Upload by Skyler