ML-MT: A study of e-health application framework by machine learning techniques

Abstract

One of the crucial challenges of the healthcare sector is to provide services efficiently so that everyone can get the care they need without having to get caught up in the administrative processes or a long waiting list. Machine learning has been used to solve various problems in healthcare and this research paper aims to increase efficiency by providing an easy-to-use platform that not only works as a bridge between the doctors and the patients but also expedites the diagnosis process with the assistance of machine learning. The proposed framework is composed of mainly two core components which are the patient management system and the doctor management system. The patient management system allows patients to request an appointment, collects and monitors patient data, and predicts the chances of the patient having heart disease, kidney disease, liver disease, cancer, and diabetes. The doctor views their appointments, patients list, and status, and provides diagnosis and treatment where the prediction from the patient management system comes in handy to make decisions faster. The doctor's assistant and hospital administrative staff work behind the scene so that appointment scheduling is running smoothly and data is properly available for the models to make predictions. Among the machine learning algorithms explored, Logistic regression had the highest accuracy values in predicting cancer (97%), diabetes (92%), and liver disease (91%). Naive Bayes was 97% accurate in predicting kidney disease and Random Forest was 94% accurate in predicting heart disease which was the best performance in the case of these two diseases.