Diabetes Prognostication – An Aptness of Machine Learning

Abstract

The expanse of machine learning has reached physical and medical sciences. The testing that were done before by physically examining a person can now be efficiently predicted by using machine learning algorithm. In today's generation diabetes is a growing health problem that has heterogeneous effects. In a report by World Health Organization (WHO), it revealed that in 2015 close to 1.6 million people died due to diabetes. The report also predicts that by 2030 diabetes will be seventh leading cause of death. In the assessment by International Diabetes Federation, more than 150 million cases of diabetes are undiagnosed. Busy lifestyle, improper food consumption and lack of physical activity on daily basis for long time has given birth to many diseases. One such disease is diabetes. It is already labelled as a Global disease. Treatment of diabetes is available but millions of people live with diabetes unknowing the fact they are suffering from it. The aim of this work is to make an user friendly, accurate and efficient low cost diabetes diagnose software, a Graphical User Interface which can predict diabetes and can be used by NGO's to diagnose people belonging to economically weaker section. This paper refers to a project which aims to classify a person's data into two classes, 'Yes' and ‘No’, based on ten factors., namely age, family history, alcoholic, smoker, etc.. In this work the accumulated result is obtained from mode of the four outputs, from four machine learning algorithms namely, SVM, KNN, ANN and Naive Bayes.