Early detection of disease can prevent fatality and even save the lives of individuals. Since many diseases may have some common symptoms, it is essential to critically analyze symptoms for the correct prediction of diseases. Machine learning influences disease prediction, analyzing numerous features with high accuracy. In our country, elderly people suffer mostly alone as every other member remains busy outside the home, so they lack proper care and constant observation. A cloud infrastructure that allows digital devices to gather, infer, and exchange health data is called the Internet of Medical Things (IoMT). As the global economy grows, so will the cost of linked healthcare. The ever-lowering cost of sensor-based technologies is the reason behind the extraordinary expansion of IoMT. This paper reviews which machine learning algorithm is most suitable for detecting non- communicable diseases in terms of precision, specificity, accuracy, and confusion matrix. It is possible to keep track of old persons by detecting disease from the early stages of symptoms. We used OHAS (Occupational Health Automated System) dataset for finding the accuracy of the disease detection system. We utilized several machine learning techniques for detecting non-communicable diseases (for example, K-Nearest Neighbor, Decision Tree, Support Vector Machine (SVM), XG-Boost, and logistic regression). This article's objective is to investigate the repercussions of using the aforementioned algorithms effectively and find out which is the best algorithm for early Disease detection. We observed that from the mentioned algorithms, XG-Boost outperforms all other algorithms and gives the best accuracy of 86.24 percent.