Prediction of diabetic retinopathy signs in diabetes and healthy subjects using debrecen datasets by comparing NB and KNN classifiers

Aim: The aim of the study is to classify the diabetic and healthy subjects using Naive Bayes (NB) and K Nearest Neighbor (KNN) machine learning algorithms from a debrecen dataset. Materials and Methods: From the UCI machine learning repository, the debrecen dataset of healthy (n=34) and diabetic (n=34) subjects were collected. The dataset consists of 19 characteristics which are considered as the inputs to classifiers. The classification of healthy and diabetic subjects was performed using the Waikato Environment for Knowledge Analysis (Weka) 3.8.5 version data mining tool. The IBM Statistical Package for the Social Science (SPSS) version 21 software was used for the statistical analysis. Results: The performance of the two classifiers were compared and found that the accuracy obtained by Naive Bayes classifier (68.68%) appears to be higher than the K Nearest Neighbor classifier. The statistical significant difference was observed between the two groups (p