Higher education institutions (HEIs) are increasingly recognized as pivotal actors in the global transition toward sustainability. However, food waste management within campus canteens remains a persistent challenge often addressed through fragmented, behavior-focused interventions rather than systemic governance. This study aims to address this gap by constructing a closed-loop, data-driven framework that bridges the divide between waste diagnosis and actionable management.
The study employs a deep learning hybrid model that integrates unstructured student feedback with multidimensional dish features to accurately diagnose waste risks at the micro-level. Building upon these diagnostic insights, a mixed-integer linear programming model is developed to optimize intervention strategies.
The empirical results demonstrate that sustainable canteen management is not a zero-sum game; the model successfully achieves a scientific equilibrium among three competing objectives: minimizing the ecological footprint, controlling operational intervention costs and safeguarding student satisfaction.
By transitioning environmental management from experience-based operations to algorithmic precision, this research provides a scalable model for precision environmental governance in HEIs. These findings directly support UN Sustainable Development Goals 12 (Responsible Consumption and Production) and 11 (Sustainable Cities), reinforcing the role of universities as living laboratories for verifying innovative sustainability solutions.
