Prediction of Diabetic Retinopathy using Novel Decision Tree Method in Comparison with Support Vector Machine Model to Improve Accuracy

Abstract

The main objective of this paper is to predict Diabetic Retinopathy (DR) using Novel Decision Tree (DT) in comparison with Support Vector Machine (SVM). Prediction of Diabetic Retinopathy is done using Novel Decision Tree (N=10) and Support Vector Machine (N=10) algorithms. Kaggle fundus image dataset which contains more than 50,000 digital retinal images is used for Diabetic Retinopathy detection. Novel Decision Tree has attained an accuracy of 92.8% whereas Support Vector Machine got only 85.2%. Both DT and SVM have a statistical significant difference of (p=0.03). Novel Decision Tree method has better performance when compared to Support Vector Machine for Diabetic Retinopathy Detection.