Which method is typically used to assess model performance on unseen data?

Cross-validation is a technique that is widely used to evaluate the performance of predictive models on unseen data. This method involves partitioning the training dataset into multiple subsets, where the model is trained on a portion of the data and validated on a different subset. By rotating through different subsets, cross-validation provides a comprehensive assessment of how well the model generalizes to new, unseen datasets.

The main advantage of cross-validation lies in its ability to mitigate overfitting, as it ensures that the model's performance is not just a reflection of its fit to the specific training data. By averaging the performance across multiple validation sets, cross-validation provides a more reliable estimate of a model’s predictive capability when applied to new data.

Other methods such as aggregation, normalization, and feature selection serve different purposes within the modeling process. Aggregation is often related to combining predictions from multiple models, normalization involves adjusting the scale of features, and feature selection is about identifying the most relevant inputs for model training. However, none of these methods specifically focus on assessing the performance of the model on unseen data as effectively as cross-validation does.