Which concept helps ensure the stability of a predictive model when outliers are present?

Regularization is a technique used in statistical modeling to prevent overfitting, particularly when outliers are present in the data. It does this by imposing a penalty on the size of the coefficients of the model. This means that the model becomes less sensitive to extreme values, allowing it to focus on the overall trend rather than being disproportionately influenced by outliers.

When a predictive model is trained on data that includes outliers, these extreme values can skew the results and lead to a model that performs poorly when applied to new data. Regularization addresses this issue by constraining the model, thereby enhancing its generalization capability. By minimizing the impact of these extreme observations, regularization helps maintain the model's stability and reliability.

In contrast, normalization typically deals with adjusting the scale of data, which can have a different effect on model performance related to distribution rather than stability. Iteration refers to the process of repeatedly refining model parameters, which may not specifically address the influence of outliers. Cross-validation is primarily a method for assessing the performance and robustness of the model on unseen data, but it does not directly mitigate the effects of outliers during the modeling process.