Machine Learning Approaches for Analyzing Physiological Data in Remote Patient Monitoring

Abstract

The surge in infectious and chronic diseases, coupled with the increasing number of independent-living elderly individuals, has made remote patient monitoring (RPM) crucial for healthcare service. Sensors technology and the Internet of Things produce numerous small robotics devices for collecting multi-modal data on relevant physiological parameters. These are subsequently analyzed to predict health conditions in time to save patients’ lives. Range-specific physiological data analysis is very crucial in RPM systems. According to the literature, deep learning effectively analyzes this data type. However, its low explainability and high resource dependency motivate us to conduct a comparative study on how various machine learning (ML) approaches can enhance RPM. This study compares the performance of decision trees, random forest, and support vector machine algorithms on two publicly available RPM medical datasets from Kaggle. We evaluate these ML techniques for RPM in terms of accuracy, precision, recall, and sensitivity. Our findings provide valuable insights into the effectiveness of ML methods in improving RPM systems.