test data

Terms from Artificial Intelligence: humans at the heart of algorithms

The glossary is being gradually proof checked, but may have typos and misspellings.

Test data is the portion of data used for testing the model created by machine learning. Typically most of the available data is used as training data with a smaller portion used as test data. This is important as models often perform better on training data than unseen data due to overlearning, that is where the model has picked up accidental patterns in the training data that are not typical of the data as a whole. Generalisaton is assessed by measuring accuracy on the test data.

Used in glossary entries: machine learning, overfitting, training data