EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Feature embedded attention based hybrid approach for athletic injury risk prediction

Mengqi Hu, Zhiqiang Song, Xiaoling Feng

Alexandria Engineering Journal · 2026-03 · 139 · 1-10

DOI: 10.1016/j.aej.2026.02.005 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Alexandria Engineering Journal, 2026-03. Verfasst von Mengqi Hu und 2 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 38 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

With the advancement of artificial intelligence driven athletes monitoring techniques, the sports analytics ecosystem gains multimodal observations for analyzing injury paradigm, providing abundant data sources for reliable athletes injury prediction. However, accurate predicting of athletes injury risk remains a unresolved challenge in sports science, where multivariate physiological, psychological, and workload factors interact in nonlinear ways. To address this challenge, this study introduces FAHC-LSTM, a feature-embedded attention-based hybrid convolutional neural network and long short-term memory framework designed to enhance reliability in athlete injury forecasting. The proposed model first applies Uniform Manifold Approximation and Projection (UMAP) to perform nonlinear dimensionality reduction and construct low-dimensional feature embeddings that capture latent dependencies among heterogeneous training and mental-status variables. A hybrid architecture combining the convolutional neural network (CNN) and the bidirectional long short term memory (LSTM) model leverages both local feature correlations and long-term temporal dependencies, with an attention mechanism strengthening the injury prediction accuracy. Experiments on the Dutch runner dataset demonstrate that FAHC-LSTM achieves 81.2% classification accuracy and 86.9% AUC-ROC, outperforming conventional machine learning (ML) and state-of-the-art deep learning (DL) baselines. Ablation studies confirm the contributions of UMAP embedding and hybrid architecture to performance stability and learnability. The results position FAHC-LSTM as a robust and generalizable solution for proactive injury monitoring, advancing personalized training management and athlete health protection.

Zitieren

APA

Hu, M., Song, Z., & Feng, X. (2026). Feature embedded attention based hybrid approach for athletic injury risk prediction. Alexandria Engineering Journal, 139, 1-10. https://doi.org/10.1016/j.aej.2026.02.005

Vancouver

Hu M, Song Z, Feng X. Feature embedded attention based hybrid approach for athletic injury risk prediction. Alexandria Engineering Journal. 2026;139:1-10. doi:10.1016/j.aej.2026.02.005

BibTeX
@article{Hu2026,
  author = {Hu, Mengqi and Song, Zhiqiang and Feng, Xiaoling},
  title = {Feature embedded attention based hybrid approach for athletic injury risk prediction},
  journal = {Alexandria Engineering Journal},
  year = {2026},
  volume = {139},
  pages = {1--10},
  doi = {10.1016/j.aej.2026.02.005}
}
Kostenlos starten →