Scholarlink

<p>Buffet restaurant supply chains face persistent challenges from demand uncertainty and ingredient perishability, generating significant food waste and stockout risks—a sustainability concern of growing urgency in hospitality operations. Conventional practices relying on heuristic ordering rules or value-based classification schemes fail to capture demand variability across multi-branch environments, revealing a critical gap in artificial intelligence (AI)-driven integrated decision support for perishable inventory management. This study proposes a three-component AI-enabled framework: (i) a LightGBM model optimized via Optuna under walk-forward validation for branch-level daily sales forecasting using calendar, weather, and temporal lag features, with forecasted sales converted into item-level raw-material requirements through historical usage-to-sales ratios; (ii) a novel Volume–Value–Consistency (VVC) segmentation scheme that extends ABC analysis by incorporating demand consistency, measured by the coefficient of variation, alongside volume and value; and (iii) VVC-guided replenishment decisions based on VVC categories. Applied to a four-branch Thai shabu-shabu buffet chain, the framework is empirically evaluated against a prior-month heuristic baseline along two complementary dimensions: spoilage cost (overstock) and service-level outcomes (understock). The LightGBM model reduced mean absolute percentage error (MAPE) by 12.6–31.2% points over lag-free baselines, while the integrated framework reduced spoilage costs by 19.4% on average, with largest reductions—up to 31.6%— observed in stable, low-variability branches. In addition, the framework reduced stockout incidence and improved demand fulfillment, indicating that waste reduction was achieved without compromising service performance. Collectively, these findings suggest that integrating predictive analytics, consistency-aware inventory segmentation, and VVC-guided replenishment decisions provides an effective decision-support approach for reducing food waste and improving inventory performance in multi-branch perishable foodservice supply chains, subject to context-specific recalibration.</p>

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

An AI-enabled decision support framework for sustainable inventory planning in buffet restaurant supply chains

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月30日

查看全文
  • Kullapapruk Piewthongngam,
  • Suphakan Homkhampad,
  • Arthit Apichottanakul

摘要

Buffet restaurant supply chains face persistent challenges from demand uncertainty and ingredient perishability, generating significant food waste and stockout risks—a sustainability concern of growing urgency in hospitality operations. Conventional practices relying on heuristic ordering rules or value-based classification schemes fail to capture demand variability across multi-branch environments, revealing a critical gap in artificial intelligence (AI)-driven integrated decision support for perishable inventory management. This study proposes a three-component AI-enabled framework: (i) a LightGBM model optimized via Optuna under walk-forward validation for branch-level daily sales forecasting using calendar, weather, and temporal lag features, with forecasted sales converted into item-level raw-material requirements through historical usage-to-sales ratios; (ii) a novel Volume–Value–Consistency (VVC) segmentation scheme that extends ABC analysis by incorporating demand consistency, measured by the coefficient of variation, alongside volume and value; and (iii) VVC-guided replenishment decisions based on VVC categories. Applied to a four-branch Thai shabu-shabu buffet chain, the framework is empirically evaluated against a prior-month heuristic baseline along two complementary dimensions: spoilage cost (overstock) and service-level outcomes (understock). The LightGBM model reduced mean absolute percentage error (MAPE) by 12.6–31.2% points over lag-free baselines, while the integrated framework reduced spoilage costs by 19.4% on average, with largest reductions—up to 31.6%— observed in stable, low-variability branches. In addition, the framework reduced stockout incidence and improved demand fulfillment, indicating that waste reduction was achieved without compromising service performance. Collectively, these findings suggest that integrating predictive analytics, consistency-aware inventory segmentation, and VVC-guided replenishment decisions provides an effective decision-support approach for reducing food waste and improving inventory performance in multi-branch perishable foodservice supply chains, subject to context-specific recalibration.