Leveraging Machine Learning for Physical Exercise Recommendation Based on Heart Rate: Older Adults Personalized Training

Healthcare 5.0 smart solutions can improve the quality of life among people, especially older adults, by tackling the negative effects of aging. The Active3 project proposes a virtual reality-based system that enables cycling exercise and a cognitive game in a multi-player VR environment to promote social inclusion among older adults via daily exercise. This paper proposes a machine learning algorithm to personalize the cycling workload for older adults to maintain safe yet stimulating cardiovascular training. It leverages historical data from the user's heart rate wearable sensor and corresponding exercise workload to predict the heart rate and suggest the proper workload accordingly. This system aims to provide personalized physical training to improve older adults' health and well-being through a customized exercise program based on the user's physiological condition in a safe environment without clinical supervision. This work integrates a linear regression model to the user's heart rate measurements and workload to develop a simple model that understands the relationship between heart rate and physical exercise workload. Cross-validation is performed on a dataset of older adults, which was collected from the previous study containing the user's heart rate, resting heart rate, and cycle ergometer workload, together with the user's age, gender, weight, and height, to analyze the effect of these features.