Sensor Based Human Activity Recognition using Machine Learning

This paper presents a sensor-based Human Activity Recognition (HAR) system using machine learning techniques to classify six physical activities cycling, resting, running, swimming, walking, and yoga. The proposed model employs a Random Forest classifier trained on features such as heart rate, step frequency, distance traveled, and step entropy. Experimental results show that the Random Forest model achieved an accuracy of 85%, outperforming a Decision Tree baseline (76.5%). Feature importance analysis revealed heart rate and step frequency as key indicators for activity classification. The system also includes a real-time activity prediction module, demonstrating its applicability in fitness tracking, health monitoring, and wearable technology. This study highlights the potential of integrating sensor data with predictive models to enable intelligent, personalized wellness and healthcare solutions.