AWS Certified AI Practitioner Practice Exam

In the context of evaluating the performance of an image classification model that is predicting plant diseases, the accuracy metric is most suitable. Accuracy measures the proportion of correctly classified instances out of the total instances. This is particularly important in classification tasks because the primary aim is to correctly identify which category each image belongs to, which in this case, are different classes of plant diseases.

Using accuracy provides a straightforward indication of how well the model performs across all classes of plant diseases. If you have a balanced dataset and the classes are roughly equal, accuracy will effectively convey the performance of the model. However, it's important to note that in cases of imbalanced datasets, other metrics like precision, recall, and F1-score may also need to be considered.

R-squared score is typically used in regression contexts to measure the proportion of variance explained by the model, making it unsuitable for a classification problem. Root mean squared error (RMSE) is another metric meant for regression tasks, assessing the average magnitude of error in predictions without considering the direction, which does not apply to classification. Learning rate is a hyperparameter that governs the speed at which a model updates during training, which is not a metric for evaluating performance after training is complete.