DIABETES DISEASE PREDICTION USING MACHINE LEARNING ALGORITHMS

Abstract

This paper deals with the prediction of 
Diabetes Disease by performing an analysis 
of five supervised machine learning 
algorithms, i.e. K-Nearest Neighbors, 
Naive Baye, Decision Tree Classifier, 
Random Forest and Support Vector 
Machine. Further, by incorporating all the 
present risk factors of the dataset, we have 
observed a stable accuracy after classifying 
and performing cross-validation. We 
managed to achieve a stable and highest 
accuracy of 76% with KNN classifier and 
remaining all other classifiers also give a 
stable accuracy of above 70%. We analyzed 
why specific Machine Learning classifiers 
do not yield stable and good accuracy by 
visualizing the training and testing accuracy 
and examining model overfitting and model 
underfitting. The main goal of this paper is 
to find the most optimal results in terms of
accuracy and computational time for 
Diabetes disease prediction.