Comparative Analysis of Ensemble Models in Machine Learning for Human Activity Recognition in Wearable Health Systems

Wearable technologies have become a crucial part of modern health monitoring systems, offering insights into physical activities and physiological states through sensors such as accelerometers, gyroscopes, and ECGs. Activity recognition, the task of identifying and classifying physical activities from sensor data, is central to applications like fitness tracking, fall detection, and chronic disease management. In this study, we compare the performance of two widely used machine learning algorithms, Random Forest and XGBoost, for activity recognition using the MHEALTH dataset, which records data from multiple wearable sensors across ۱۲ distinct activities. Our results show that Random Forest, with class weight adjustments, outperforms XGBoost, achieving an accuracy of ۹۴.۷۲%, while XGBoost achieved ۹۳.۰۴%. This study highlights the importance of class imbalance handling and contributes to improving real-time health monitoring systems by demonstrating the effectiveness of class-weight adjustment in ensemble models for human activity recognition.

کلیدواژه ها:

Human Activity Recognition ، Wearable Sensors ، Machine Learning ، Personalized Health Monitoring ، Random Forest ، XGBoost ، Activity Classification ، Class Imbalance ، Real-time Monitoring