Which dataset is typically used to fit the parameters of a predictive model?

• A. Scoring
• B. Testing
• C. Validation
• D. Training

Answer: (D)

The dataset used to fit the parameters is the training data. During training, the model adjusts its internal parameters—like weights or coefficients—based on the input-output pairs in this set, with the goal of minimizing a loss function and capturing the patterns in the data. This is why it’s the one that gets used for fitting: it provides the actual examples the model learns from.

Validation data are kept separate to tune hyperparameters and monitor overfitting, giving a sense of how the model might perform on unseen data during development. Testing data are held aside even more strictly and used only to estimate how well the final model will generalize after training and hyperparameters have been chosen. Scoring is a way to quantify performance, but it doesn’t drive the parameter updates.