Graph Neural Network and Cloud-Based Intelligent Recommendation System for Student Physical Fitness

Abstract

Physical fitness significantly supports students' health, academic success, and overall development in modern education. Many educational institutions now adopt advanced technologies to monitor physical activities and suggest personalized training programs based on individual needs. However, most existing systems rely on simple models that cannot fully capture the complex relationships among various fitness indicators, such as heart rate, endurance, and flexibility. These systems also face challenges in processing large datasets efficiently and delivering real-time, personalized feedback to diverse student groups. The Graph-Based Intelligent Cloud Framework for Student Fitness (GICF-SF) has been developed to address these challenges. GICF-SF utilizes Graph Neural Networks (GNNs) to analyze complex interactions within physical fitness data, enabling a more accurate understanding and recommendation. Additionally, Cloud Computing supports fast data storage, processing, and real-time response, providing scalability for multiple users across different locations.GICF-SF integrates machine learning and cloud technologies to deliver tailored training suggestions by learning from each student's unique fitness profile. The cloud infrastructure allows the framework to serve schools, fitness centers, and online platforms efficiently without performance.The proposed GICF-SF model uses a 3-layer Spatial-Temporal Graph Convolutional Network (ST-GCN) with ReLU activation and dropout. The student fitness dataset (10,421 records) was split into 70% training, 20% testing, and 10% validation sets. Evaluation was performed using precision, recall, F1-score, MAE, and RMSE metrics under 5-fold cross-validation. Results show that GICF-SF improved recommendation accuracy by 12.8% and reduced training time by 17.3% over traditional methods.