You have an Azure Machine Learning workspace. You plan to use the workspace to set up automated machine learning training for an image classification model. You need to choose the primary metric to…

explained_variance is a regression metric that quantifies how much of the variance in a continuous target variable the model explains and cannot be applied to classification label outputs.

mean_absolute_error measures the average magnitude of prediction errors for continuous numeric outputs and is only valid for regression tasks, not classification.

Image classification is a multi-class classification problem, and accuracy measures the fraction of images correctly labeled by the model. Azure AutoML supports accuracy as a primary metric for classification tasks and uses it to compare and select the best model during training iterations.

median_absolute_error is a robust regression metric based on the median of absolute errors and is not applicable to categorical class predictions.