부산대학교 도서관

The prevention of visual loss in diabetic individuals depends critically on the early identification of diabetic retinopathy (DR). For the purpose of early detection DR, this research proposes a hybrid soft computing technique. The method uses many algorithms to create a classification model that is effective in spotting early indications of DR in retinal images. The aim of this research is to increase the DR classification's precision and dependability, especially for early-stage DR images. With the use of the Multivariate Minimum Redundancy-Maximum Relevance (MRMR) approach, the dimensions of the characteristics are decreased. The modified water cycle algorithm is used to optimize the SVM hyper-parameters for the RBF kernel function. The findings show that the algorithm under consideration has an effective strategy when compared to similar methods in terms of accuracy.