HealthCloud: A system for monitoring health status of heart patients using machine learning and cloud computing

• Published in: Internet of things (Amsterdam. Online) Vol. 17; p. 100485
• Main Authors: Desai, Forum, Chowdhury, Deepraj, Kaur, Rupinder, Peeters, Marloes, Arya, Rajesh Chand, Wander, Gurpreet Singh, Gill, Sukhpal Singh, Buyya, Rajkumar
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.03.2022
• Subjects: Cloud computing; Heart disease prediction; Machine learning; Smart healthcare; Cloud computing; Heart disease prediction; Smart healthcare; Machine learning
• ISSN: 2542-6605; 2542-6605
• DOI: 10.1016/j.iot.2021.100485

In the context of the global health crisis of 2020, the tendency of many people to self-diagnose at home virtually, prior to any physical interaction with medical professionals, has been increased. Existing self-diagnosis systems include those accessible via the Internet, which involve entering one’s symptoms. Several other methods do exist, for example, people read medical blogs or notes, which are often wrongly interpreted by them and they arrive at a completely different assumption regarding the cause of their symptoms. In this paper, a system called HealthCloud is proposed, for monitoring health status of heart patients using machine learning and cloud computing. This study aims to offer the ‘best of both worlds’, by combining the information required for the person to understand the disease in sufficient detail, with an accurate prediction as to whether they may have (in this case) heart disease or not. The presence of heart disease is predicted using machine learning algorithms such as Support Vector Machine, K-Nearest Neighbours, Neural Networks, Logistic Regression and Gradient Boosting Trees. This paper evaluates these machine learning algorithms to obtain the most accurate model, in compliance with Quality of Service (QoS) parameters. The performance of these machine learning models is measured and compared using the metrics such as Accuracy, Sensitivity (Recall), Specificity, AUC scores, Execution Time, Latency, and Memory Usage. For better establishment of the results, these machine learning algorithms have been cross validated with 5-fold cross validation technique. With an accuracy rate of 85.96%, it has been found that Logistic Regression is the most responsive and accurate model amongst those models assessed. The Precision, Recall, Cross Validation mean and AUC Score for this model were 95.83%, 76.67%, 81.68% and 96% respectively. The algorithm and the mobile application were tested on Google Cloud Firebase with existing user inputs from the dataset, as well as with unseen new data. The use of this system can assist patients, both in reaching self-diagnosis decisions and in monitoring their health.