Fusion of Classical ML and LSTM Architectures for Enhanced Human Activity Recognition in Modern Healthcare Technology

The increasing need for effective Human Activity recognition (HAR) is driven by applications in healthcare, impacting user experience and safety. Traditional machine learning approaches are efficient and interpretable. But ML models struggle with temporal dependencies in sequential data. In contrast, Long Short-Term Memory (LSTM) network surpasses other models in capturing such complex patterns. This study proposes a LSTM model to improve HAR performance in healthcare tools, achieving 97% accuracy. The architecture enables accurate real- time recognition of patient movements and activities and detects early anomalies like falls. It also continuously monitors the mobility patterns of the patient. The proposed model is suitable for wearable healthcare devices and scalable remote patient monitoring systems as it balances computational efficiency and temporal modeling. This proposed architecture shows superiority to other models in terms of accuracy, robustness and adaptability, contributing practical approaches to enhance patient safety, and care quality in modern healthcare.