Supply chain demand prediction is a critical challenge in modern logistics and inventory management. This paper presents a comprehensive analysis of machine learning regression models for accurate demand forecasting, focusing on their predictive performance and interpretability. We evaluate traditional linear models, ensemble models, gradient boosting machines, and neural networks, comparing their effectiveness on supply chain datasets. To enhance model transparency, we employ Local Interpretable Model-agnostic Explanations (LIME) to interpret predictions and identify key drivers of demand. Our results demonstrate that linear models and ensemble methods, such as linear regression, ridge regression, and lightGBM, achieved the lowest MSE of 75.03, 75.04, and 85.20, respectively, and the highest R2 score of 0.99 for all of them. The integration of LIME offers actionable insights into model predictions, enabling stakeholders to understand and trust the forecasting process. This study highlights the importance of combining predictive accuracy with interpretability in supply chain applications, providing a framework for deploying machine learning models in real-world logistics scenarios.