A model that memorizes is a model that fails

Overfitting is when a model learns the training data too well — including its noise — and then performs poorly on new data. The cure is honest evaluation.

Train / validation / test split

Never judge a model on the data it learned from. Hold out a test set it has never seen.

Pick the right metric

• Regression: MAE, RMSE, R².
• Classification: accuracy is a trap when classes are imbalanced — use precision, recall, F1, and the confusion matrix.

Keep it honest

• Use cross-validation for a stable estimate.
• Watch the gap between training and test scores — a big gap means overfitting.
• Regularize, simplify, or get more data to close it.

Takeaway: the goal isn't a high training score — it's generalization to the real world.