Improved Accuracy of Predicting Diabetic Retinopathy in Type 2 Diabetes Mellitus using Genetic Algorithm

Introduction: diabetes mellitus is a prevalent disease and its late diagnosis leads to dangerous complications and even death. one of the serious complications of this disease is diabetic retinopathy, the leading cause of blindness in the developed countries. because of slowly progressive nature and lack of symptoms in the early stages of the disease, it is essential to predict the probability of developing diabetic retinopathy promptly to implement the appropriate therapy. methods: our dataset contains 29 extracted features from 310 patients with types 2 diabetic disease, 155 patients of whom sufferred from diabetic retinopathy. the patients were selected randomly from motahari clinic in shiraz, iran between 2013 and 2014. first, the genetic algorithm, (ga) as a feature selection process, was implemented to select the most informative features (high-risk factors) for prediction of diabetic retinopathy. then, three well-known classifiers including k-nearest neighbors (knn), support vector machine (svm), and decision tree (dt) were applied to the optimized dataset for classification of the two mentioned groups. results: our finding showed that ga selected 13 factors for better prediction of diabetic retinopathy; these factors were the duration of the disease, history of stroke, family history, cardiac diseases, diabetic neuropathy, ldl, hdl, blood pressure, urine albumin, 2hppg, hba1c, fbs, and age. given the selected risk factors, the classification accuracy was obtained 69.35%, 81.29% and 96.13% by svm, dt, and knn, respectively. our results showed that knn had the highest accuracy in the prediction of diabetic retinopathy compared to svm and dt, and the difference between knn and the other algorithms was statistically significant. conclusion: the proposed approach was compared and contrasted with recently reported methods, and it was shown that a considerably enhanced performance was achieved. this research may aid healthcare professionals to determine and individualize the required eye screening interval for a given patient.