Design cross-validation; explain bias–variance

Give a precise definition of cross-validation, then contrast k-fold, stratified k-fold, leave-one-out, nested CV, and time-series rolling or blocked CV. For a dataset that has both temporal ordering and class imbalance, propose an evaluation setup that prevents leakage while still yielding stable estimates; explain your fold construction and choice of metric. Quantitatively explain the bias–variance tradeoff: how model complexity, training set size, and regularization shift bias and variance; how cross-validation error curves reveal underfitting or overfitting; and which levers—features, model class, regularization, data augmentation—you would adjust when variance is high versus when bias is high.

Overview: This question tests knowledge of cross-validation approaches and the bias–variance tradeoff, including competence in sound model evaluation, experimental design for time-ordered and class-imbalanced data, and quantitative reading of cross-validation error curves.