Using Association Rules for interaction detection/feature creation
Interactions allow one to see the combined effect of more than one variable. Unfortunately, interactions are not automatically calculated by many algorithms. The Association Rules created here are intended to find interactions between nominal, ordinal, and flag variables in the data. In this recipe we will create 10 new interactions to use as model inputs using the A Priori Association Rules node. This recipe builds from the Selecting variables using single-antecedent Association Rules recipe from Chapter 2, Data Preparation – Select, including using the same target variable: the
TARGET_D quintile between $20 and $200.
This recipe uses the datafile