Enhancing Health Monitoring using Efficient Hyperparameter Optimization

Abstract

Nowadays, healthcare problems among elders have been increasing at an unprecedented rate, and every year, more than a quarter of the elderly people face weakening injuries such as unexpected falls, etc. resulting in broken bones and serious injuries in some cases. Sometimes, these injuries may go unnoticed, and the resulting health consequences can have a considerable negative impact on their quality of life. Constant surveillance by trained professionals is impossible owing to the expense and effort. The detection of physical activities by different sensors and recognition processes is a key topic of research in wireless systems, smartphones and mobile computing. Sensors document and keep track of the patient's movements, to report immediately when any irregularity is found, thus saving a variety of resources. Multiple types of sensors and devices are needed for activity identification of a person's various behaviours that record or sense human actions. This work intends to gather relevant insights from data gathered from sensors and use it to categorize various human actions with machine learning using appropriate feature selection and hyperparameter tuning, and then compare the implemented models based on their performance. Understanding human behaviour is very useful in the healthcare industry, particularly in the areas of rehabilitation, elder care assistance, and cognitive impairment.