Enhancing Diabetic Retinopathy Detection: A Novel Ensemble Random Support Vector Gradient Model

Globally, Diabetic Retinopathy (DR) is a major complication disease affecting millions of people, and an important cause of visual impairment for individuals with diabetics. Thus, this paper introduces a novel Ensemble Random Supportvector Gradient (ERSG) model using four important phases including data collection, pre-processing, feature extractionandclassificaation. The data collection step gathers images from three datasets namely the Diagnosis of DR, the DR-G, and the DR datasets. The next phase is to pre-process the input image, this includes three methods such as normalization, contrast enhancement, and resizing. Gray-Level Co-Occurrence Matrix (GLCM) performs the feature extraction by generating several features including contrast, correlation, energy, homogeneity, and entropy. The final phase is to detect and classify DR by applying weighted ensemble of machine learning methods such as Adaptive Support Vector, Random Forest, and eXtreme Gradient Boosting (XGBoost). A weighted Average Ensemble is employed to integrate the predictions of each classifier. Finally, comprehensive experimental analyses are conducted to affirm the detection ability of the proposed model, where several estimation keys are employed to measure its performance. The ERSG attains higher performance with higher accuracy and precision of 98.9% and 97.3% than existing DR methods.