Predicted Multi-Chronic Disease by Supervised Machine Learning Algorithms: Performance and Evaluation

Abstract

Current environmental conditions and human lifestyles have resulted in the emergence of numerous diseases. The medical field generates an enormous amount of new data each year for remote monitoring of patients. Due to increased data growth in the medical and healthcare industries, accurate medical data analysis has been advantageous to early patient care. However, physicians often face challenges in accurately diagnosing diseases in patients far from hospitals. Therefore, utilizing remote patient systems (telemedicine systems) due to the complexities associated with their chronic conditions. On the other hand, predicting illness is also a challenging task. Thus, data extracted from heterogeneous, fast-flowing, and reliable sources is crucial for decision-making and disease prediction. This research paper aims to utilize supervised Machine Learning (ML) techniques to predict chronic diseases such as heart and hypertension based on the patient’s features or symptoms by analyzing patient data collected by sensors and sources enabled by the Internet of Medical Things (IoMT). Supervised ML technology in Hadoop and Spark environments is employed to guarantee that this classification accurately identifies individuals with chronic illnesses. The methods are evaluated using 55,680 patient records to discover the proper match between the data set and the final disease-predicted result. The results demonstrate that the proposed procedure employing the Decision Tree (DT) algorithm is 94% accurate, and DT outperforms the other four ML algorithms. This includes the Support Vector Machine (SVM), a Naive Bayes (NB) model, Random Forest (RF), and Logistic Regression (LR) in terms of both performance and accuracy metrics (precision, recall, and F-score).