<p>Individuals should engage in exercise programs tailored to their health status and physiological characteristics to support overall health. This study presents a personalized health management and activity monitoring system that uses an Improved Bat Algorithm-optimized Long Short-Term Memory (IBA-LSTM) network. The system employs wearable IoT sensors to collect real-time physiological and activity data, including heart rate, blood oxygen saturation, temperature, and activity levels. For more accurate modelling, the signals are preprocessed and placed into a time series. To improve prediction accuracy and stability, an enhanced bat-inspired metaheuristic optimization approach is employed. We tested the new model on real-world and benchmark datasets and demonstrated improved predictive performance, computational efficiency, and robustness than regular LSTM and other baseline models. The IBA-LSTM model was able to obtain an accuracy of 94.2%, which is more than what Genetic Algorithms and Particle Swarm Optimization can do. The system is scalable and low-latency, supporting deployment on mobile and wearable devices while enabling proactive health insights and personalized exercise recommendations. Overall, the proposed framework demonstrates improved predictive performance and computational efficiency, indicating its suitability for real-time smart preventive healthcare applications.</p>