Eye disorders have become one of the leading causes of vision loss around the globe. This highlights the importance of early accurate diagnosis of these disorders. Most of the current methods for diagnosing eye diseases use manual examination of retinal images and depend primarily on experienced ophthalmologists. These manual methods can take a lot of time to examine each eye(s) and can also be highly subjective in their determination of a diagnosis.In this project, we will develop an automated framework that uses machine-learning techniques to automatically detect and classify multiple eye disorders from retinal images.We only used traditional machine-learning methods in this work and did not consider using deep-learning models. The proposed framework includes systematic image preprocessing for standardising input images, automated feature extraction using feature extraction methods, and dataset balancing using the synthetic minority oversampling technique (SMOTE) to adjust for class imbalance in each of the datasets used for training and testing our classification system. The proposed framework includes the use of supervised machine-learning algorithms for the classification of all images in each class. The performance of the proposed machine-learning framework was evaluated thoroughly using various performance metrics including confusion matrices, ROC curves, precision–recall analysis, learning curve graphs, PCA visualisation, and cross-validation stability analysis. The results from the experimental evaluations indicated that the proposed machine-learning framework can generate very high classification accuracy and good generalisability for the various types of diseases evaluated using the proposed machine-learning classification system. Overall, the experimental results show that the proposed machine-learning framework can provide good quality decision support tools to use for early eye disorder diagnosis in clinical settings.