Inference, estimation, and decision-making from data
A model can score almost perfectly on the rows it was trained on and still be worthless on the next batch of real data. High training accuracy only proves the model can fit examples it has already seen; it says nothing about whether it learned a pattern that carries over to new cases. To find out, you split the data you have into pieces that each do a different job.
🔒 This is a Pro lesson — the interactive figure, worked examples, quiz and practice open with Pro access.
▶ Train, Validation & Test Sets