Cross-Validation
Definition
Cross-validation evaluates models over multiple repeated training/validation splits.
In cross-validation, training and validation splits are formed multiple times according to a suitable scheme and the results are evaluated together. k-fold cross-validation is one variant; group, time, and data dependencies determine the appropriate split.
Simply explained
Like conducting an exam multiple times with changing withheld tasks.
Why it matters in practice
Relevant for robust model selection with limited datasets.
Example from business
Several models are compared with grouped 5-fold CV before the final test set is opened.
Limits and pitfalls
- Include test set in CV; randomly shuffle time series; distribute data from the same customer across folds; report only averages.
Origin & history
Cross-validation is an established statistical resampling method; Kohavi empirically compared variants in 1995.
Use in mid-sized companies
- Possible application in SMEs: secure model and hyperparameter selection with appropriate fold rules.
Frequently asked questions
Does CV replace a final test set?
Not always. Which CV fits? That depends on data structure and goal.
Sources
- Kohavi (1995) · Kohavi (1995)
Sources checked on 10.10.2026.
Editorially reviewed on: