cross-validation
UK[ˌkrɒsˌvælɪˈdeɪʃən]US[ˌkrɔːsˌvælɪˈdeɪʃən]
n
A technique in machine learning and statistics for assessing how the results of a statistical analysis will generalize to an independent dataset, by partitioning data into subsets, training the model on some subsets, and validating it on the remaining subsets.
n
The process of checking or proving the validity or accuracy of something by using an alternative method or source.
Morpheme Breakdown
cross
validation
cross
across
validation
act of making valid
Etymology
The term is a modern compound, primarily emerging in the mid-20th century within statistical and computational sciences. Its first element, 'cross', draws from the Latin 'crux', which originally denoted a physical stake or cross. The sense evolved to signify intersecting lines and, by metaphorical extension, the concept of traversing or reciprocal action. The second element, 'validation', originates from the Latin 'validus' (strong, powerful), which in legal and logical contexts came to mean "sound" or "effective." The suffix '-ation' solidifies it as a process. Thus, 'cross-validation' literally means "the process of establishing strength or effectiveness through reciprocal or intersecting checks," perfectly capturing its technical method of using multiple data partitions to test a model's robustness.
Analysis
Structure: cross (across) + validation (act of making valid)
- cross: From Old English 'cros', from Old Norse 'kross', from Old Irish 'cros', from Latin 'crux, crucis' (stake, cross). Functional role: A prefix or combining form meaning "across," "transverse," or "reciprocal."
- validation: From the base word 'valid' (from Latin 'validus' meaning strong, effective) + the noun-forming suffix '-ation' (from Latin '-atio', indicating an action or process). Functional role: The core noun denoting the action or process of establishing validity.
Examples
We used 10-fold cross-validation to ensure our model was not overfitting the training data.
The researcher employed cross-validation techniques to verify the initial experimental findings.
A key step in developing a reliable algorithm is rigorous cross-validation against an independent dataset.