Thailand is experiencing a significant demographic shift toward an aging population, leading to increased healthcare demands. This research aims to address personalized health management for the elderly by developing a reinforcement learning system based on the FITT (Frequency, Intensity, Time, and Type) standard. The system monitors and recommends personalized activities to improve health outcomes. Data on elderly activities were collected using wearable devices and mobile apps, capturing the full FITT components. After preprocessing, the K-Nearest Neighbors (KNN) algorithm was used for classification, with hyperparameter tuning via Grid Search to optimize performance. Additional models, including Decision Trees, Random Forests, SVM, and Logistic Regression, were implemented and compared. Ensemble methods further enhanced predictive accuracy. The KNN model demonstrated high classification accuracy, while the reinforcement learning-based recommendation engine effectively adapted to user-specific data, offering comprehensive and personalized activity plans. Challenges such as data variability and user compliance highlight the need for further innovation, including improved sensor technologies and strategies for engagement. In conclusion, the integration of the FITT framework and advanced machine learning algorithms resulted in a robust system that improves health outcomes and user satisfaction. This technology-driven approach provides a promising tool for caregivers and healthcare providers, laying the foundation for enhanced elderly care through personalized exercise recommendations. Future work will focus on refining the system with real-world testing and enhancing user adherence.