Ristivalidointi

Ristivalidointi, or cross-validation, is a statistical resampling method used to assess how a predictive model generalizes to an independent dataset. The approach partitions the available data into training and validation sets multiple times and aggregates the resulting performance measures to obtain a more reliable estimate than a single train-test split.

The most common form is k-fold cross-validation. The data are split into k folds of roughly equal size. For each iteration, the model is trained on k−1 folds and evaluated on the remaining fold. The process is repeated until each fold has served as a validation set, and the performance scores are averaged. Leave-one-out cross-validation is a special case with k equal to the number of observations.

Variants include stratified cross-validation, which preserves class proportions in each fold and is useful for imbalanced datasets. Nested cross-validation uses an inner loop for hyperparameter tuning and an outer loop for unbiased performance estimation, aiding model selection. Repeated cross-validation performs multiple random splits to reduce variability in the estimate.

Advantages include efficient use of data, reduced variance in the estimated performance, and broad applicability to small datasets. Disadvantages include higher computational cost, potential leakage if data are not independent, and limitations for time-series or non-iid data.

Ristivalidointi is widely used to compare models, select hyperparameters, and report performance estimates in machine learning studies. Choosing an appropriate variant depends on the data characteristics, the goal of the analysis, and the available computational resources.