Definition
Cross-validation is a resampling technique used to evaluate machine learning models on a limited data sample. It involves partitioning the data into multiple subsets, training the model on some subsets, and then testing it on the remaining subset. This process is repeated multiple times with different partitions to ensure a more robust assessment of the model's performance and generalization ability. Cross-validation helps prevent overfitting by providing a more reliable estimate of how the model will perform on unseen data. For instance, a data scientist might use k-fold cross-validation to assess a model's accuracy in predicting customer churn. This differs from a simple train-test split, which only evaluates the model once on a single split of the data. Concepts related to model evaluation and validation frequently appear on AI and data science certification exams.