(Machine|Statistical) Learning - (Target|Learned|Outcome|Dependent|Response) (Attribute|Variable) (Y|DV)
About
An (outcome|dependent) variable is a measure that we want to predict.
There is two type of outcome variables:
- <math>Y</math> : the original score collected
- <math>\hat{Y}</math> : the predicted score (or estimator) from the equation. The hat means “estimated” from the training data.
A dependent variable (DV) is measured during the experimentation in contrast with independent variable (IV) that describes:
- the characteristics of the group (such as name)
- and/of the experimentation (such as the steps: before, after).
The target Attribute in your build data of a supervised model is the attribute that you try to predict.
The target attribute is also called:
- Outcome measurement (Y)
- Outcome variable (Y)
- Output variable
- Label
- Dependent variable (DV),
- Response variable. (It's a statistic term when groups respond to a treatment)
- Responding variable
- Regressand
- Explained variable
- Experimental variable
- Output variable
- Criterion Variable
- Unobserved variable
- Predicted variable
- Measured variable
When the target attribute is:
- continuous, it's a regression problem and the target is a number
In the test data, it will contain values with known outcomes in order to measure the performance of the model.
In the scoring data, it will contains the results when the model is applied.
Clustering, feature extraction, association, and anomaly detection models do not use a target because they are unsupervised function.
When the target is predicted, we get an predicted target and you can find the following notations in formulas.
- <math>Y</math> is the original target score from the training data (ie collected)
- <math>\hat{Y}</math> is the predicted score from the model.
Articles Related
Terminology
- classification: The learned attribute is categorical (“nominal”)
- regression: The learned attribute is continuous numeric value.
Tools
Weka
