Real-Time Student Behavior Monitoring System Based on Edge–Cloud Collaboration and LSTM Modeling

To meet the demand for intelligent governance under the "Three-All Education" framework in universities, this study presents the design and implementation of a real-time student behavior monitoring system. Built upon an edge–cloud collaborative architecture, the system integrates multi-source data—such as access control logs, classroom attendance, power usage, library entries, and card transactions. Edge nodes powered by Jetson Nano handle local data preprocessing, feature extraction, and initial classification, significantly reducing latency and bandwidth usage. At its core, an LSTM (Long Short-Term Memory) model is used to capture temporal behavior patterns for anomaly detection. When tested on real campus data, the model achieved a detection accuracy of 94.3% and an F1 score above 0.91, outperforming Random Forest and SVM benchmarks. Deployed at a technical university in Jiangsu Province covering more than 4,000 students, the system processes over 15,000 samples daily, with end-to-end latency under 2 seconds and reliable stability. These results demonstrate the system’s practical viability and strong potential for supporting student management, risk warning, and intelligent governance in smart campus settings.