Genome-based Prediction of Diabetes Using Machine Learning

Abstract

Genome based prediction is widely popular in public health care sector due to its ability to predict the onset of common diseases, thus helping patients in early diagnosis and recovery. In this paper, we have created a simple Keras Neural Network with one hidden layer having 5 hidden nodes. A sequential model is used and after it trains on the data it predicts the likelihood of diabetes on the testing data. In the end, a graph is plotted to prove that more the diabetes pedigree function, more likely it is for a patient to get diabetes. This proves that ancestor history plays an important role when it comes to prediction of diabetes.