An AI-powered knowledge graph framework for injury prediction in professional sports

The emergence of Industry 4.0 has transformed professional sports into data-intensive ecosystems, yet injury prediction remains a persistent challenge due to fragmented information sources and limited interpretability of existing models. Addressing this gap, we present a Knowledge Graph based Injury Prediction Framework powered by a Dual-Attention BiEncoder (DABIE) for context-aware triplet extraction. The platform merges a variety of data sources, including biomechanical sensor outputs, medical reports, task records, and unstructured match comments, into a single knowledge graph that holds semantically rich information about players, events, and damage causes. DABIE employs two types of attention over phrase context and entity-pair semantics to get the exact triplets that are relevant to the injury. These triplets are stored and organized in the knowledge graph, which makes it possible to learn more through graph embeddings. The system encodes items and relations in low-dimensional vector spaces by combining translational and bilinear embedding models with margin-based ranking loss. Injury prediction is therefore defined as a link prediction task, which estimates a player’s association with future injury events. To ensure explainability, predictions are complemented with graph reasoning queries, allowing practitioners to trace causal pathways such as workload–injury correlations. Testing shows that the suggested method is more interpretable and performs better than state-of-the-art relation extraction and prediction models in terms of accuracy, recall, F1-score, and semantic correctness. Modern sports coaches, analysts, and medical personnel can benefit from proactive, clear, and practical information provided by these findings, which provide the groundwork for cyber-physical injury prevention systems.