Authors: Israa Odeh; Mouhammd Alkasassbeh; Mohammad Almseidin
Addresses: Department of Computer Science, Princess Sumaya University for Technology, Amman, Jordan ' Department of Computer Science, Princess Sumaya University for Technology, Amman, Jordan ' Department of Computer Science, Aqaba University of Technology, Aqaba, Jordan
Abstract: Diabetic retinopathy (DR) is one of the leading causes of vision loss in diabetics worldwide. DR is a microvascular disease that affects the eye's retina, leading to vascular blockage and thus disrupting the primary source of nourishment for the retinal tissue. This paper merged a considerable range of known classification algorithms into a sophisticated diagnostic model and a new multilayer perceptron (MLP)-based classifier. The latter achieved the highest accuracy rates among all other classifiers, including the previously introduced ensemble model. Four sub-datasets were generated by WrapperSubsetEval. and InfoGainEval. Feature selection algorithms achieved 75.8% and 70.7% accuracy rates on the best performing sub dataset (InfoGainEval. top 5) by the ensemble model and the new MLP-based classifier. The results express the splendid performance of the partial dataset with the MLP artificial neural network, which contributes significantly to a less complex classification process than the original full Messidor dataset.
Keywords: diabetic retinopathy; ensemble learning; machine learning; neural networks; MLP; multilayer perceptron.
International Journal of Bioinformatics Research and Applications, 2022 Vol.18 No.4, pp.300 - 317
Received: 19 Mar 2022
Accepted: 01 Jun 2022
Published online: 03 Nov 2022 *