This paper presents a method to classify diabetic retinopathy using fundus images. In our study we categorize the disease into two classes: diabetic retinopathy non-proliferative and diabetic retinopathy proliferative. The method reduces the dimensionality of the images and find features using the statistical method of principal component analysis (PCA). Then, we classify the images using decision trees, the naive Bayes classifier, neural networks, k-nearest neighbors and support vector machines. The experimental results show that the naive Bayes classifier obtains the best results with 73.4% of accuracy using a data set of 151 images and testing with different resolutions.