Semi-Supervised Learning Frameworks for Smart Campus Analytics

Abstract

The proliferation of Internet of Things (IoT) technologies and smart infrastructure in educational environments has given rise to the concept of smart campuses, which leverage data from various sensors and systems to optimize operations, security, and academic services. However, the vast volume of heterogeneous data collected poses significant challenges in terms of labeling and effective utilization for analytics. In this context, semi-supervised learning (SSL) emerges as a promising approach that combines a small amount of labeled data with a large pool of unlabeled data to improve learning accuracy.

This manuscript presents a comprehensive exploration of semi-supervised learning frameworks for smart campus analytics, emphasizing how these models can bridge the gap between data abundance and limited human annotation. It begins by detailing the architecture of smart campuses and the data sources involved—such as surveillance systems, RFID-based attendance tracking, energy usage meters, and student interaction logs. Following a literature review of contemporary SSL techniques—including pseudo-labeling, graph-based methods, and consistency regularization—the study proposes a novel hybrid SSL framework combining graph convolutional networks (GCNs) with self-training to handle smart campus datasets effectively.

Methodologically, a simulation study is conducted using synthetic and real-world datasets from an anonymized university smart campus environment, focusing on three primary use cases: student performance prediction, facility usage optimization, and anomaly detection in energy consumption. Statistical analysis, including precision, recall, and F1-score comparisons with fully supervised and unsupervised models, is provided in tabular form. Results demonstrate that the proposed semi-supervised approach achieves superior performance with limited labeled data, ensuring efficient and scalable campus analytics.

The conclusion underscores the transformative potential of SSL in educational analytics and suggests directions for integrating federated learning and privacy-preserving methods to safeguard sensitive student data. This research contributes to the growing field of AI-driven campus management and offers practical implications for educational institutions aiming to enhance operational intelligence with minimal annotation cost.