Diabetes is a condition that is spreading quickly over the globe. It happens when the body cannot utilise enough insulin or when the pancreas does not create enough of it. Diabetes Retinopathy (DR) and/or blindness is one of the main issues that diabetic individuals face. Diabetes patients are becoming more numerous, which also results in an increase of DR data. Therefore, the use of data mining (DM) techniques becomes important to extract the useful information and undiscovered knowledge from the data. Data Mining has become crucial in DR detection since it can contribute to the society's overall health. The present study focuses on the quick, straightforward, and simple prediction of diabetic retinopathy using the physical records of the patients. Information is extracted from these records, which are in the form of numerical values, using DM techniques. Numerous classifiers have been used in the dataset for this DR prediction, including logistic regression, KNN, SVM, bagged trees, and boosted trees. Two distinct cross validations are applied to the data to find the best features and guard against overfitting. Dataset is consisted of the records of 900 diabetes patients. This novel work confirms that bagged trees and KNN are effective classifiers for DR prediction utilising physical health records. According to the results of the experiment, boosted trees with a 10% hold-out validation have the best classification accuracy at 90.1%. KNN has also resulted in 88.9% accuracy, which is noteworthy

Diabetic Retinopathy, Data Mining, Logistic regression, SVM, KNN, Boosted tree, Bagged tree, Cross Validation