Machine Learning Applied in Demand Forecasting and Supply Planning
Zhang, Yi (2024)
2024
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2024052214146
https://urn.fi/URN:NBN:fi:amk-2024052214146
Tiivistelmä
Effective demand forecasting and supply planning are crucial elements in supply chain management. Inaccurate demand information in the supply chain often leads to suboptimal decision-making, resulting in inventory imbalances and customer dissatisfaction. This study aims to address these challenges by leveraging ML algorithms and models to enhance demand forecasting accuracy and optimize supply chain operations for the case company.
This study employed an applied action research approach to diagnose the case company's challenges and offer possible solutions. Qualitative methods, including interviews, meetings, and internal document analysis, were primarily utilized for data collection, supplemented by some quantitative data use for model development. Four algorithms: Linear Regression, Decision Tree, Recurrent Neural Network, and Support Vector Machine (Vandeput 2023) were employed to build ML models by using data extracted from the company's weekly demand reports. After data processing, feature engineering, training, testing, and validation, Linear Regression emerged as the most appropriate algorithm based on both ML metrics and internal evaluation.
The outcome of the thesis is a proposed ML-based approach how to reduce excess stock and improve supply shortage that is recommended for integration into the company's existing demand forecasting and supply planning processes to assist decision-making.
Although the existing research discusses ML applications in demand forecasting and supply chain management, this study contributes by providing a practical implementation tailored to a real-world company context. Through this study, it aims to pave the way for similar solutions for the case company and wider, in this field.
This study employed an applied action research approach to diagnose the case company's challenges and offer possible solutions. Qualitative methods, including interviews, meetings, and internal document analysis, were primarily utilized for data collection, supplemented by some quantitative data use for model development. Four algorithms: Linear Regression, Decision Tree, Recurrent Neural Network, and Support Vector Machine (Vandeput 2023) were employed to build ML models by using data extracted from the company's weekly demand reports. After data processing, feature engineering, training, testing, and validation, Linear Regression emerged as the most appropriate algorithm based on both ML metrics and internal evaluation.
The outcome of the thesis is a proposed ML-based approach how to reduce excess stock and improve supply shortage that is recommended for integration into the company's existing demand forecasting and supply planning processes to assist decision-making.
Although the existing research discusses ML applications in demand forecasting and supply chain management, this study contributes by providing a practical implementation tailored to a real-world company context. Through this study, it aims to pave the way for similar solutions for the case company and wider, in this field.