Predictor operators
compute a prediction model based on
historic data in the input table.
The historic data contains
known values for a target column and
the prediction model establishes a relationship between the other
input fields and the target column. This model can then be used on
other data to predict an unknown value where the target column value
is unknown.
For example, an insurance company has data about
customers who
allowed their insurance to lapse and those who did not (historic data).
The company can use this information to identify such customers in
the future.
Depending on the type of selected target column,
the predictor
operator uses:
- The classification mining function for predicting
categorical
target fields, for example, a field that contains values like Yes
or No. Using the Classification mining function, you can select one
of the following classification algorithms:
- Tree classification
- Naive Bayes classification
- Logistic Regression classification
- The regression function for predicting numerical target
fields,
for example, total revenue for a customer with values like $1234.12.
Using the regression function, you can select one of the following
regression algorithms in the predictor operator properties:
- Transform
regression
- Linear regression
- Polynomial regression
- Radial Basis Function (RBF)
Tip: The
Predictor operator creates a prediction model
over historic data that can be used for scoring with the scorer operator
on future data. The Predict column operator should be used instead
if the input table contains both, known and unknown values for the
target column. The Predict column operator creates a new table that
contains predictions for the unknown values.