4 Regression Free Online FlashCards

What does regression predict?

Regression is a supervised learning method that uses features to predict a numeric target.

What does ordinary least squares minimize?

Ordinary least squares chooses coefficients that minimize the sum of squared differences between actual and predicted targets.

Why keep a test set separate from training?

Fit the model using only the training set. Keep the test set separate to estimate performance on examples the model did not train on.

What does mean absolute error (MAE) measure?

Mean absolute error is the average absolute prediction error. It is expressed in the target's units.

What do the sign of a residual tell you?

A residual is the actual target minus the predicted target. A positive residual means the model predicted too low; a negative residual means it predicted too high.

What does an $R^2$ of $0$ mean on a test set?

An $R^2$ of $0$ means the model performs as well as predicting the test-set mean.

What do fit and predict do in a regression workflow?

fit learns the model parameters from training data; predict uses the fitted model to produce targets for new feature rows.

In regression, what are a feature and a target?

A feature is an input value used for prediction; the target is the numeric value the model learns to predict.

What does the $250 slope mean in the home-price example?

In the example, the model associates each additional square foot with a $250 increase in predicted price.

What does a linear model's intercept represent?

The intercept is the prediction when the feature is zero. It may have little practical meaning if zero is outside the useful data range.

How should you interpret a coefficient in a multiple-feature model?

It describes the model's predicted change for that feature while holding the other included features fixed; it does not prove causation.

How does RMSE treat large errors compared with MAE?

Root mean squared error is expressed in the target's units and penalizes large errors more strongly than MAE.