Deep learning-based dynamic prediction of physical energy expenditure and training intensity optimization system for university sports training

Accurate estimation of physical energy expenditure (EE) is essential for optimizing training intensity and preventing overexertion in university sports programs. Existing EE prediction models fail to account for individual physiological modifications and the dynamic relationship between multimodal physiological signals and energy consumption. To overcome these limitations, this research develops a deep learning (DL)-based dynamic prediction and training intensity optimization system for university athletes. The research uses a multi-modal dataset (2500 university athletes) containing dynamic signals (tri-axial acceleration, heart rate (HR), and electro cardio gram (ECG)) and static physiological parameters Body Mass Index (BMI), body-fat percentage, resting HR, and resting dynamic oxygen uptake (VO2)). Data pre-processing removes noise and normalizes signal scales. The feature extraction captures temporal and spatial features from static and dynamic signals. Feature fusion integrates the two modalities for personalized representation. The fused data are fed into the proposed Teaching-Learning Fused Intelligent Convo Memory Network (TL-ICMN), which combines Convolutional Neural Networks (CNN) for local temporal feature learning, an Attention mechanism (AM) for attribute selection, an Improved Long Short-Term Memory (ILSTM) network for long-term dependency modelling, and Teaching-Learning-Based Optimization (TLBO) for hyper-parameter tuning and performance enhancement. The model using Python 3.11 achieves strong predictive performance with lower error rates, Training time (0.5s) and a high correlation coefficient (R2) of 0.955, representative excellent agreement between predicted and actual EE. The proposed personalized TL-ICMN framework dynamically predicts EE with high precision and optimizes training intensity, by offering an intelligent decision-support tool for university sports training and performance management.