Predictive AI and Machine Learning Models for Performance, Injury, and Health in Sport and Active Living

Machine learning and predictive artificial intelligence are increasingly used to forecast sports performance, assess injury risk, monitor athlete health, and support active-living interventions. Although many studies report high predictive accuracy on benchmark or retrospectively collected datasets, concerns remain regarding overfitting, data leakage, limited sample diversity, inadequate reporting of learning dynamics, and the absence of external or prospective validation. These challenges raise important questions about whether reported performance reflects genuine predictive capacity and whether models can generalize across athletes, sports, populations, and real-world environments. Recent research has advanced model development, wearable sensing, multimodal data integration, and explainable artificial intelligence, but methodological practices remain inconsistent. Efforts to address these limitations increasingly emphasize transparent reporting, robust benchmarking, reproducible workflows, and collaboration between sports scientists, clinicians, data scientists, and practitioners. Nevertheless, important gaps persist in understanding model behavior during training, establishing reliable standards for validation and reporting, protecting sensitive performance data, and translating predictive systems into safe, equitable, and usable workflows.This Research Topic aims to promote methodological rigor, transparency, reproducibility, and practical relevance in predictive AI and machine learning research for sport and active living. It will encourage studies that examine learning dynamics, identify and mitigate overfitting and data leakage, evaluate model robustness through external or cross-context validation, and clarify the conditions under which predictive systems can support sports science, athletic training, sports medicine, and health promotion. Contributions should clearly report data processing, model development, training behavior, evaluation procedures, uncertainty, and limitations. The Research Topic also seeks to address how explainability, privacy, fairness, and data governance influence the responsible deployment of predictive technologies. Guiding questions include how model learning dynamics are reported and interpreted, which strategies support generalizability, how studies advance transparency and reproducibility, and what barriers must be overcome for practical implementation.To gather further insights into the development, evaluation, and translation of reliable predictive AI systems for sport and active living, this Research Topic will focus on studies involving athletic, recreational, clinical, and population-level contexts, while prioritizing methodological evidence over claims based solely on predictive accuracy. Relevant themes include, but are not limited to:Reporting, visualization, and interpretation of machine-learning learning dynamics in sports and health predictionStrategies for preventing overfitting, data leakage, selection bias, and inflated performance estimatesExternal, prospective, and cross-domain validation of predictive modelsBenchmarking models across sports, populations, competition levels, and environmental settingsGeneralizability and robustness across elite athletes, recreational exercisers, and diverse participant groupsExplainability, interpretability, uncertainty quantification, and transparent reporting standardsDataset curation, annotation quality, documentation, sharing, and reproducible data workflowsIntegration of wearable, sensor, performance, clinical, and contextual dataPractical deployment in athletic training, sports medicine, injury prevention, and health promotionPrivacy, fairness, security, and data governance in performance-monitoring systemsHuman–AI collaboration and interdisciplinary approaches combining sports science and AI developmentStandardized datasets, open challenges, evaluation protocols, and reproducible benchmarks for sports AI and machine learningBarriers to implementation, including usability, ethics, regulation, cost, and practitioner trustDevelopment of user-adaptive models and out-of-distribution data handling. Topic editors: Milan Toma, Jeawoog Lee, Enrique Garcia Ceja Source: https://www.frontiersin.org/research-topics/84334/predictive-ai-and-machine-learning-models-for-performance-injury-and-health-in-sport-and-active-living