DataTerminology.comNEW
Also known as: Target
An output variable is the target a supervised model learns to predict from input features.
DESCRIPTION
In supervised learning, the output variable (target or label) is the quantity a model attempts to predict. It can be numeric for regression or categorical for classification. Training pairs input features with the output to learn patterns that generalize to new data. Clear definition and quality of the output variable are crucial for valid evaluation and deployment. In experiments, it corresponds to the dependent variable affected by the independent variables.
ALTERNATIVE DEFINITIONS
Wikipedia
The dependent variable is the variable being tested and measured in a scientific experiment.
Frames the output variable as the dependent variable in scientific experiments.
Google Machine Learning Crash Course
A label is the thing we’re predicting—the y variable in y=f(x).
Uses the ML term “label” to describe the predicted target.
RELATED TERMS
Predictor Variable Supervised Learning Regression Analysis Classification Oracle Loss Function Model Evaluation Inference
COMPETING TERMS
Input Variable
KNOWLEDGE AREAS
TARGET AUDIENCE
Data Scientist AI Engineer MLOps Engineer Data Analyst Data Modeler
EXTERNAL SOURCES
- Wikipedia - Dependent and independent variables
- scikit-learn - Glossary entry for target variable
- Google Developers - Evaluation metrics relate predictions to labels
TOP QUESTIONS
What does the output variable represent in machine learning?
It represents the quantity a supervised model aims to predict based on input features. It is crucial for defining the learning objective and evaluation metrics.
Category: Concept
How is an output variable different in regression versus classification?
In regression, it is continuous; in classification, it is discrete class labels. The choice determines suitable loss functions and metrics.
Category: Misunderstanding
Why is output variable quality important?
Noisy or mislabeled outputs mislead the model and degrade performance. High‑quality labels are essential for reliable training and evaluation.
Category: Risk
How is the output variable used during evaluation?
Predictions are compared to the true output variable using metrics like accuracy, F1, RMSE, or MAE, depending on the task.
Category: Testing
Can the output variable be engineered or transformed?
Yes. Transformations (e.g., log for skewed targets) can stabilize variance or model multiplicative effects, but complicate interpretation.
Category: Best practices
What is another term for output variable in ML?
Target or label are common synonyms, especially in libraries like scikit‑learn and TensorFlow.
Category: Alternatives
How do class imbalance issues affect the output variable?
Imbalanced classes bias training towards majority labels. Techniques like class weighting or resampling mitigate this.
Category: Risks
Where do output variables come from in practice?
They often come from historical outcomes, human annotations, or derived business events. Provenance and governance are critical.
Category: Process
What is a common mistake with output variables?
Using proxies that don’t reflect the real objective (e.g., clicks as a proxy for satisfaction) leads to misaligned optimization.
Category: Mistakes