Scholarlink

This study evaluates the effectiveness of machine learning models in improving order quantity forecasting for Vietnamese coffee production. Using a two-year dataset comprising 851 products, the research focuses on key variables, including purchasing costs, delivery costs, holding costs, market demand, and order variance. The performance of XGBoost, Random Forest and ARIMA is compared using the mean absolute error and mean absolute percentage error to assess predictive accuracy. The Importance Score model is used to identify the three most critical products, which are analyzed using an Economic Order Quantity framework. This approach integrates purchasing, delivery, and holding costs with market demand to determine optimal order quantities. The results reveal that Random Forest performs best in high-variability datasets, while XGBoost excels in stable demand scenarios. ARIMA, although effective for capturing trends, struggles with nonlinear patterns. The study underscores the value of machine learning in addressing the complexities of agricultural supply chains. By reducing forecast errors and improving responsiveness, these models provide a robust framework for optimizing decision-making processes. The results demonstrate the potential of machine learning to improve operational efficiency and profitability, offering practical insights for the coffee production industry.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Leveraging Machine Learning for Accurate Order Quantity Forecasting: A Case Study of Vietnam’s Coffee Production Industry

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Nguyen Nguyen Khoi,
  • Kha Pham Dong,
  • Dung Tran Anh,
  • Anh Hoang

摘要

This study evaluates the effectiveness of machine learning models in improving order quantity forecasting for Vietnamese coffee production. Using a two-year dataset comprising 851 products, the research focuses on key variables, including purchasing costs, delivery costs, holding costs, market demand, and order variance. The performance of XGBoost, Random Forest and ARIMA is compared using the mean absolute error and mean absolute percentage error to assess predictive accuracy. The Importance Score model is used to identify the three most critical products, which are analyzed using an Economic Order Quantity framework. This approach integrates purchasing, delivery, and holding costs with market demand to determine optimal order quantities. The results reveal that Random Forest performs best in high-variability datasets, while XGBoost excels in stable demand scenarios. ARIMA, although effective for capturing trends, struggles with nonlinear patterns. The study underscores the value of machine learning in addressing the complexities of agricultural supply chains. By reducing forecast errors and improving responsiveness, these models provide a robust framework for optimizing decision-making processes. The results demonstrate the potential of machine learning to improve operational efficiency and profitability, offering practical insights for the coffee production industry.