Introduction
Creates new variables containing respondent level predicted probabilities of each response from a machine learning model or ensemble.
Requirements
A predictive model run using linear discriminant analysis, random forest, support vector machine or some other machine learning algorithm in Q.
Method
Example
Once the model has been run, save the probabilities by clicking Probabilities of Each Response button in the Save Variables(s) section of the object inspector:
This can also be run by selecting Automate > Browse Online Library > Machine Learning > Save Variables > Probabilities of Each Response from the menus.
A table of probabilities for each outcome category is generated automatically:
Technical details
Uses the Probabilities method from the R package flipData. Unlike the defaults in most R packages, this implementation assigns probabilities for observations not in the estimation sample.
Next