Performance Analysis of Machine Learning Techniques for Diabetic Retinopathy Detection

Abstract

Diabetic retinopathy is an eye disease that affects the light sensitive area of retina. It does not give any sign at initially but later it become very difficult to cure it. So it is very essential to detect it at initial stage for this various computer aided software is designed using machine learning techniques. This paper analyzed the most widely used machine learning techniques used for this disease detection i.e. SVM, KNN, decision tree, random forest, logistic regression neural network, naive bayes and deep learning architecture i.e. CNN, VGG16, ResNet50, EfficientNetB0, InceptionV3, CNN and SVM . Performance of these techniques is analyzed using five different datasets. AUC, CA, F1 score, precision, and recall are used for evaluation purpose.