Early Stage Diabetes Prediction Using Machine Learning Methods

Diabetes is a common disease that incurable and fatal. Millions of people worldwide have diabetes it directly affects people’s lives. Early diagnosis helps reduce the effects improve life quality patients, but in case live with for years before getting diagnosed. can be done by applying machine learning methods on existing data patients. In this way, quickly get diagnosed without taking glucose screening test or any blood test. Answering simple question set would enough to determine if person diabetic has risk being diabetic. proposed study, determination performed techniques. scope, publicly available dataset, which includes 16 features are collected from 520 people, was used create predictive models. Eight were individually over dataset. The results each model validated using 10 fold cross validation schema. Addition accuracy metric, confusion matrix based other performance metrics; precision, recall f1 score, also reported. All created models resulted high scores. minimum score measured as 88.85% one basic techniques, Naive Bayes. highest rate 99.04%, obtained dimensional convolutional neural network model. designed Convolutional Neural Network scores metrics 100.00%, 98.63% 99.31% scores, respectively. These findings indicate 1D CNN utilized patients asking only several questions