In the context of the digital transformation of higher education, the Smart Campus concept has emerged as a strategic direction for the development of university infrastructure. One of the key elements of this concept is the integration of Internet of Things (IoT) sensors, enabling systematic data collection on behavioral, spatial, and physiological parameters of students. However, existing research in the field of learning analytics is predominantly focused on analyzing LMS logs and does not incorporate real-time data from the physical campus environment, thereby limiting the ability to detect academic risks at early stages. This study aims to develop and experimentally validate an integrated architecture for real-time monitoring of learning activities within a Smart Campus environment, based on IoT sensor data streams and predictive analytics methods. The proposed approach integrates data on attendance, student mobility, microclimate parameters, attention levels, and digital activity into a unified analytical platform. Within the framework of this research, a multi-layer data processing model was developed, including a sensor data acquisition layer, streaming aggregation layer, machine-learning analytics layer, and decision-support visualization layer. Predictive models were implemented using Random Forest, XGBoost, and LSTM neural networks. The effectiveness of the models was evaluated using Accuracy, F1-score, ROC-AUC, and RMSE metrics. The results demonstrate an improvement in the early prediction accuracy of academic underperformance by 18-23% compared to traditional LMS-based approaches. The proposed architecture enables the implementation of early warning systems and adaptive management of the educational environment. The findings confirm that the integration of IoT data and predictive analytics methods forms a new paradigm for managing the quality of higher education in digital universities, facilitating the transition from reactive to proactive models of academic support.