Testing Your AIConclusion

K-fold cross-validation squeezes a fair score out of a small dataset. It costs k trainings instead of one, but it rescues you from the luckcostspeed of a single split, and it lets every row help with both learning and checking.

Keep the order of jobs straight: cross-validate on the training data to choose the model and its settings, then open the locked test set once for the final, honest score.