Model Prediction

Alessandra Corvonato

2023-04-14

Initial Situation and Goal

In addition to the 'Cornerstone' core methods of fitting data by a linear regression it is possible to use a a model building function from 'CornerstoneR' to model data. Afterwards, the model can be used to make predictions for other datasets.

How do we use the method 'modelPredict' in 'Cornerstone' from 'CornerstoneR'?

Use Fitted Models for Predictions

The following conditions are required in order to properly use the ‘modelPredict’ function from CornerstoneR:

• an already fitted Logistic Regression, Decision Tree, Random Forest or Gaussian Process Regression Model outputted using the corresponding CornerstoneR method,
• data records that the fitted model did not see beforehand, but show the same structure (same predictors recorded) as the input data from the model build.

You can find the detailed user guides with examples for fitting and predicting a model in the respective user guide for Logistic Regression, Decision Tree, Random Forest and Gaussian Process Regression, section “Use fitted model name for predictions”.

Options in the Script Variables Dialog

Starting from the 'R' analysis object 'modelPredict', you can find the 'Script Variables' dialog via the menu 'R Script' \(\rightarrow\) 'Script Variables'. The following dialog appears.

Model Prediction: Script Variables

Here, you can choose if you wish to output computed columns as such. That means, if your input data contains computed columns, they will show up in the outputted 'Predictions' dataset as static columns. If you check this box in the script variables, the computed columns will be outputted with their formula. Make sure that the computation of these columns is still valid and not using any columns that were not passed to the CornerstoneR routine.