{"entity":{"id":"train-test-discipline","kind":"term","name":"Train, validation and test split discipline","aka":["train/validation/test split","training set","validation set","test set","held-out test set","held-out test split","80/20 split","hold-out set"],"tldr":"Data is divided into a training set the model learns from, a validation set used to choose settings, and a test set touched once at the end; using the test set to make choices turns it into a second validation set.","summary":"In machine learning the input data used to build a model is divided into a training set, a validation set for tuning and a test set for the final, unbiased evaluation (Wikipedia). The discipline is behavioural: the test set is looked at once, patients (not samples) are the unit of splitting so one patient's slides do not appear on both sides, and the split is decided before modelling. A held-out 80/20 split is the minimum; an external cohort from another institution is the real test.","asOf":"2026-09-24","wikipedia":"https://en.wikipedia.org/wiki/Training,_validation,_and_test_data_sets","links":[{"label":"Wikipedia","url":"https://en.wikipedia.org/wiki/Training,_validation,_and_test_data_sets"}],"tags":["cansim-terms"],"related":["cancer-ai-vocabulary"],"cancers":[],"sections":[],"technologies":[],"targets":[],"drugs":[],"companies":[],"institutions":[],"pathways":[],"terms":["cross-validation","data-leakage","external-validation","pre-registered-experiment"],"trials":[],"people":[],"bottlenecks":[],"keyPapers":[],"journals":[],"dependsOn":[],"notes":["Listed in the CanSim terms map 1.0.0 (docs/onco/terms.json, generated 2026-09-24), CC BY 4.0, attribution: CanSim project, an open, public-data-first cancer foundation-model programme; CanSim page path /terms/train-validation-test."],"provenance":{"editedBy":"OnCo CanSim terms wave (Wikipedia summaries, standards and project pages, GDC and FDA pages, Europe PMC)","editedOn":"2026-09-24","note":"CanSim terms map 1.0.0 (docs/onco/terms.json, generated 2026-09-24), CC BY 4.0, attribution: CanSim project, an open, public-data-first cancer foundation-model programme"},"category":"Methods and models"},"route":"/terms/train-test-discipline/","neighbours":{"term":[{"id":"cancer-ai-vocabulary","kind":"term","name":"Cancer AI vocabulary (CanSim terms map)","route":"/terms/cancer-ai-vocabulary/"},{"id":"cross-validation","kind":"term","name":"Cross-validation and stratified k-fold","route":"/terms/cross-validation/"},{"id":"data-leakage","kind":"term","name":"Data leakage in model evaluation","route":"/terms/data-leakage/"},{"id":"external-validation","kind":"term","name":"External validation","route":"/terms/external-validation/"},{"id":"pre-registered-experiment","kind":"term","name":"Pre-registered experiment","route":"/terms/pre-registered-experiment/"}]}}