International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 01 | Jan 2023
p-ISSN: 2395-0072
www.irjet.net
Managing Uncertainty in Fashion Supply Chains: An AI-Based Analysis of Demand Variability and Forecast Precision Utkarsh Mittal Manager, Machine Learning and Automation, Gap Inc. Student Department of Computer Science, Stanford University ---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract -
the authors name can be used along with the reference number in the running text. The order of reference in the running text should match with the list of references at the end of the paper.
Purpose: This study aims to enhance demand forecasting accuracy in supply chain management using advanced techniques like genetic algorithm-optimized deep neural networks and fuzzy clustering. Accurate demand prediction enables informed decision-making for inventory planning, risk mitigation, and operational efficiency.
Supply chain volatility presents significant challenges for businesses, especially in terms of its impact on demand forecasting accuracy (O'Neal, 2021). This directly affects inventory planning, risk mitigation and overall operational efficiency. While the complexity of this issue is acknowledged (Wan & Evers, 2011), prior academic works have exhibited certain limitations.
Methodology: The study adopts an empirical approach utilizing sales datasets to develop machine learning and deep learning models that effectively capture fluctuations in demand across products with varying levels of volatility. The products are classified into categories using fuzzy clustering based on demand variability metrics.
Although trend forecasting has shown to increase supply chain stability (Miyaoka & Hausman, 2008), joint forecasting models employed by Aviv (2002) rely on limited data and statistical methods. Advanced AI/ML applications for enhancing resilience have been underexplored. Despite attempts to reduce inventory fluctuations and costs using moving average techniques (Yuan et al., 2020), the bullwhip effect persists in distribution networks.
Results: The optimized deep neural network model, fine-tuned by a genetic algorithm, achieved the highest precision with under 3% mean absolute percentage error in forecasting demand variations, outperforming methods like linear regression and Temporal Fusion Transformer networks.
While machine learning has been recognized for improving supply chain efficiency (Aamer et al., 2020), its incorporation specifically for pharmaceutical demand forecasting has been recent (Yani & Aamer, 2022). Moreover, the impact of enhanced predictions on decentralized networks requires careful evaluation (Miyaoka & Hausman, 2008). Although information sharing between parallel supply chains boosts forecasting accuracy (Zhang & Zhao, 2009), corresponding gains for individual members remains unclear.
Practical Implications: The findings demonstrate the vital role of AI/ML in enhancing supply chain resilience through improved demand forecasting. By proactively adapting to demand changes, businesses can optimize inventory and production planning, leading to increased profitability, agility, and sustainability. Originality: To the best of the author's knowledge, this is the first study incorporating genetic algorithm-optimized deep learning and fuzzy clustering to categorize retail products based on demand volatility signatures, thereby significantly improving forecast accuracy.
This study aims to address prevailing gaps by adopting an empirical approach to develop optimized machine learning models that capture complex demand fluctuations across retail products. The core objectives are:
Key Words: Machine Learning, Supply Chain Risk, Deep Learning, Temporal Fusion Transformer (TFT), Artificial Neural Network (ANN), Genetic Algorithm (GA)
1. INTRODUCTION This document is template. We ask that authors follow some simple guidelines. In essence, we ask you to make your paper look exactly like this document. The easiest way to do this is simply to download the template and replace(copy-paste) the content with your own material. Number the reference items consecutively in square brackets (e.g. [1]). However,
© 2024, IRJET
|
Impact Factor value: 8.226
|
To classify products into categories based on demand variability and lifecycle stage using fuzzy clustering.
To evaluate various state-of-the-art forecasting techniques in predicting future sales for differently volatile product groups.
To determine the feasibility of tailoring forecasting models based on the unique volatility signatures identified through clustering.
ISO 9001:2008 Certified Journal
|
Page 476