Page 73 - eBook_Proceedings of the International Conference on Digital Manufacturing V1

Leveraging Machine Learning to Enhance Supply Chain Agility and Strategic
Operational Excellence

offered a broad view of sales, inventory, and logistics, while the
industry dataset provided granular, real-time transactional
records. Significant preprocessing was required to address data
sparsity, missing values, and outliers, following best practices
recommended by Chen et al. (2021) for preparing industrial
datasets.

Model selection was driven by algorithmic robustness and
predictive capability. Random Forest and Extreme Gradient
Boosting (XGBoost) were chosen, due to their superior
performance in complex, high-dimensional data environments, as
highlighted by Ma and Sun (2020). The model development
comprised of three stages; preprocessing, model training and
evaluation as shown in Figure 13.

Pre - Model Evaluation
Processing Training

Figure 13: Steps for ML Model Development

Both models were trained using a standard train-test split
methodology, and performance was evaluated via Mean Absolute
Error (MAE), Root Mean Square Error (RMSE), and the
coefficient of determination (R²), ensuring rigorous model
validation in line with standards proposed by Shinde and Shah
(2023). The implementation phase of machine learning model
consists of the following stages as shown in Figure 14.

Figure 14: Steps for ML Model Implementation

57