The rapid integration of artificial intelligence, the Internet of Things, and big-data analytics has transformed how basketball is taught, trained, and evaluated. However, conventional basketball instruction is still dominated by experience-based, one-size-fits-all programs that struggle to address heterogeneous athlete profiles, dynamic loads, and individual injury risks. To bridge this gap, this paper proposes a personalized-training-oriented data analysis and Intelligent Decision Support System (IDSS) for basketball teaching. The system fuses multimodal data — Inertial Measurement Unit (IMU) signals, video-based pose estimation, technical-tactical statistics, physiological metrics, and subjective wellness questionnaires — within a four-layer architecture comprising data acquisition, feature engineering, intelligent decision, and visualization. A hybrid recommendation engine that couples Improved Neural Collaborative Filtering with content-based athlete profiling is introduced for personalized training prescription, and an XGBoost–SHAP model is employed for interpretable performance prediction and lower-extremity injury risk estimation. A workload-balanced optimization formulation is presented to schedule training sessions under fatigue, skill, and time constraints. Empirical evaluation on a curated dataset of 120 collegiate basketball players over 18 weeks demonstrates that the proposed system improves shooting accuracy by 4.35%, reduces non-contact injury incidence by 28.6%, and yields a recommendation precision of 0.873 — outperforming conventional baselines. The findings indicate that data-driven, explainable IDSS frameworks can serve as a credible companion to coaches and physical-education teachers in modern basketball pedagogy.
Yu Lei (Fri,) studied this question.