Deep Learning-Based Adaptive Recommendation and Multi-Level Security Architecture for Smart Canteen Management Systems

Abstract

In modern smart canteens, accurate personalized recommendations and robust security are essential for operational efficiency and user satisfaction. Traditional systems often face low accuracy, delayed response, and weak data protection. This study proposes an e-Cantong smart canteen system that integrates deep neural networks (DNNs) for feature extraction, reinforcement learning for adaptive path optimization, and a real-time feedback mechanism to dynamically adjust recommendations to changing user demands and environments. For security, a layered framework combining AES encryption, user authentication, and role-based access control is designed to ensure privacy and stability under high concurrency. Experiments on cafeteria operation records and user behavior datasets demonstrate 91.3% recommendation accuracy and 1.5-second inference latency, with stable performance in large-scale scenarios. The innovation lies in unifying adaptive recommendation and multi-level security, offering a practical path for intelligent canteen management that enhances efficiency, resilience, and user experience in complex environments.