Background: The surge in health data from wearable devices enables personalized fitness recommendations, but privacy concerns and opaque machine learning models limit user trust. Federated Learning (FL) offers privacy-preserving model training, while Explainable AI (XAI) techniques like SHAP and LIME improve transparency. However, combining FL and XAI for secure and interpretable personalized fitness recommendations is underexplored.

Objectives: This study develops an explainable, privacy-preserving FL framework for personalized fitness recommendations by (1) protecting user data with FL, (2) enhancing transparency with XAI, and (3) delivering trustworthy fitness guidance.

Methods: A classification-based model was trained using the Flower FL framework across decentralized datasets. SHAP and LIME provided global and user-specific explanations. The model’s performance was evaluated using standard metrics and compared with centralized models.

Results: The FL model matched centralized performance while preserving privacy and offering personalized recommendations. SHAP and LIME improved trust through explainability. Challenges like computational overhead and data heterogeneity impacted performance.

Conclusions: The proposed FL framework effectively delivers privacy-preserving, interpretable fitness recommendations. Future work should optimize efficiency and scalability for broader real-world use.