Diabetic retinopathy is an eye-related neurological disorder, the diabetic patient eye damaged by blood vessel in the retina area of the eye. Computational methodology is a proper way for detecting and predicting the diabetic retinopathy disease. The aim is to identify and detect the Diabetic Retinopathy, so this present work focusses on detection of Diabetic Retinopathy. This work proposed the novel WMD-MSVM -Weighted Mahalanobis Distance based Multiclass Support Vector Machine oriented; upon Diabetic Retinopathy diagnosis system for the purpose of feature selection, also ROI extraction method being utilized to fetch features from Diabetic Retinopathy images.