A label is the outcome a statistical or machine-learning model predicts or explains. It is also called a target, response, dependent, or output variable. section 5.0 section 14.6
The label determines the modeling task. Regular-season wins provide a continuous label, while postseason qualification provides a binary label. section 5.0 section 14.6
Labels appear in several model forms:
- Multiple linear regression estimates how predictors collectively relate to variation in a continuous label. section 5.0
- Regression trees predict continuous label values by partitioning observations with if-else rules. section 5.0
- Logistic regression models a binary label using one or more predictor variables. section 14.6
NBA modeling questions
A model can use hustle statistics such as blocked shots, deflections, and loose balls recovered as predictors of regular-season wins. A separate model can use offensive or defensive scoring statistics to predict postseason qualification. section 5.0 section 14.6
