ABSTRACT
In recent times, machine learning and artificial intelligence are very commonly used methods in real-time classification and prediction problems. But when it comes to medicine the domain has always been questioned about its abilities, because medicine and health care involve and deal with human lives hence the level of precision and accuracy should be beyond the most experienced doctors at this time. In this paper, a methodology using machine learning is explored to identify the coherent groups of diabetes patients using unsupervised learning. A dimensional reduction technique is initially used to reduce the dimension before grouping the instances of the patients. A cluster validation technique called silhouette coefficient is used to validate the cluster. The results indicate that all clusters discovered by using unsupervised learning are valid.
