Diabetic retinopathy (DR) is a complication of diabetes that affects one in four people with the disease. Retinal degeneration occurs as a result. DR is a leading cause of preventable blindness, & early detection through automated screening is important to ensuring the health of a patient’s eyes. Preventing a significant loss of vision requires prompt medical attention, ideally from an ophthalmologist. Given the recent advancements in machine learning (ML) applied to healthcare, intelligent systems may prove effective in the early diagnosis of DR. This research seeks to identify the best method for screening diabetic retinopathy patients using datasets such as Messidor-2 datasets by comparing eight widely used classification algorithms (Inception-V3, DR2Net, ResNet50, IncRes-v2, CNN, SVM, RetNet-10, and ELM with CNN-SVD). A number of metrics, like F1-score, Precision, Recall, Accuracy, and AUC-ROC, are utilized to evaluate and contrast different algorithms. From the obtained results, it is observed that proposed RetNet-10 model outperforms other algorithms, having an accuracy of 99.46% along with a good AUC-ROC score of 99.67%. In addition, Inception-V3 obtains an accuracy of 94.59% on Messidor-2 dataset, which is quite low as the dataset is small. The simulation results show how classification algorithms perform differently, and this can be a decisive factor in which approach to take.