Correctly splitting data into train, validation, and test sets is fundamental to building models that actually work in production.

The Three Sets

• Training set (60–70%): Used to fit the model
• Validation set (15–20%): Used to tune hyperparameters and compare models
• Test set (15–20%): Used once at the end to report final performance

The Golden Rule

Never touch the test set until you're done building your model. Using test data to guide decisions is data leakage — your reported performance won't hold in production.

K-Fold Cross Validation

When data is limited, rotate through k different train/validation splits — more reliable estimate of true performance.

Reference:

scikit-learn cross-validation guide

https://scikit-learn.org/stable/modules/cross_validation.html