Table of contents

Linear node

Linear regression is a common statistical technique for classifying records based on the values of numeric input fields. Linear regression fits a straight line or surface that minimizes the discrepancies between predicted and actual output values.

Requirements. Only numeric fields can be used in a linear regression model. You must have exactly one target field (with the role set to Target) and one or more predictors (with the role set to Input). Fields with a role of Both or None are ignored, as are non-numeric fields. (If necessary, non-numeric fields can be recoded using a Derive node.)

Strengths. Linear regression models are relatively simple and give an easily interpreted mathematical formula for generating predictions. Because linear regression is a long-established statistical procedure, the properties of these models are well understood. Linear models are also typically very fast to train. The Linear node provides methods for automatic field selection in order to eliminate nonsignificant input fields from the equation.

Tip: In cases where the target field is categorical rather than a continuous range, such as yes/no or churn/don't churn, logistic regression can be used as an alternative. Logistic regression also provides support for non-numeric inputs, removing the need to recode these fields.
Note: When first creating a flow, you select which runtime to use. By default, flows use the IBM SPSS Modeler runtime. If you want to use native Spark algorithms instead of SPSS algorithms, select the Spark runtime. Properties for this node will vary depending on which runtime option you choose.