Training Data

Training data is the set of examples used to fit a machine learning model. Depending on the method, it may be labeled pairs of inputs and correct outputs, or raw unlabeled text, images, or logs.

Why it matters

A model can only learn patterns present in its data. Gaps, errors, duplicates, and skew carry straight into predictions, which is why bias in AI often traces back to what was collected. Legal questions about licensing and privacy also begin here.

Example

Practitioners usually split data into three parts:

train      70%  fit the model
validation 15%  tune settings
test       15%  final unbiased check

The test portion must never influence training, or the measured accuracy will be inflated.

Common mistakes

• Data leakage: letting information from the test set or the future slip into training.
• Assuming more data always helps even when it is low quality or unrepresentative.
• Failing to document where data came from and whether you may use it.

Related

• Dataset
• Data cleaning
• Overfitting
• Supervised learning

This page is general information written by our editorial team and updated when documentation changes. Servers and providers differ, so test changes safely first. See our editorial policy and disclaimer.