What does "cross-validation" refer to in model evaluation?

Cross-validation refers to a robust technique used in model evaluation that assesses how the results of a statistical analysis will generalize to an independent dataset. The idea is to partition the original data into subsets, using some of these subsets to train the model and the remaining subsets to test it. This method helps to ensure that the model performs well not just on the data it was trained on but also on new, unseen data.

By effectively assessing the model's performance, cross-validation helps to mitigate the risk of overfitting, which occurs when a model learns the noise in the training data instead of the actual underlying patterns. Through various forms of cross-validation, such as k-fold or leave-one-out, the same dataset is used in multiple rounds of training and testing, which provides a more reliable estimate of the model's ability to generalize.

The other options do not accurately describe cross-validation. Ignoring outliers is a data preprocessing technique unrelated to model evaluation. Bypassing model training entirely would not allow for any evaluation, as a model requires training before it can be assessed. Removing bias from data is more about data preparation and often requires different strategies than the evaluation process that cross-validation provides.