Diabetic Condition Classification using Supervised and Unsupervised Machine Learning Techniques: A Use Case for HealthTech Industry

Diabetes is a very underestimated and, in many cases, an ignored disease. People do not understand the seriousness of diabetes. Many people think diabetes is just something because of which one cannot consume sweets. This, however, is not true. Diabetes is a very serious problem where symptoms can range from being very thirsty all the time to even death. There are many cases where diabetic people had to get their leg, or some other limb amputated because of it. Thus, people have to be made cognizant of diabetes’ seriousness. This article dives deep into this disease and lists the different types of diabetes, and the way people can know if they are diabetic. Also, there is an in-depth literature review of research papers ranging from 2002-2020. Furthermore, a deep dive in the project of diabetes prediction is present. The techniques used to classify into ‘diabetic’ or ‘non-diabetic’ are Random Forest, Logistic Regression and K Nearest Neighbor algorithms which all belong to the supervised machine learning algorithms. Logistic Regression uses sigmoid function to classify into different classes, Random Forest approach uses multiple Decision Trees and the majority decision of them all for the final output. Moreover, K Nearest Neighbor technique uses the concept of Euclidean Distance to decide which class is suitable for any new coming datapoint.