Methods and formulas for Discover Key Predictors with TreeNet® Regression

Note

This command is available with the Predictive Analytics Module. Click here for more information about how to activate the module.

The predictor elimination analysis first builds a model with all the predictors. The analysis uses that model to calculate the importance scores for all the predictors. The next steps depend on whether the analysis eliminates unimportant predictors or important predictors.

Unimportant predictors
The analysis begins by sorting the predictors in order by their importance scores. If the analysis eliminates unimportant predictors, then the analysis builds sequential models by removing the least important predictors in the sorted list. More specifically, at each model step, the analysis removes any predictors with importance scores of 0 plus the specified number of least important predictors. The analysis builds a model with this subset of the predictors. The analysis uses that model to recalculate the importance values for all the predictors that remain in the analysis. The steps of predictor elimination and model construction repeat for the maximum number of elimination steps.
Important predictors
The analysis begins by sorting the predictors in order by their importance scores. If the analysis eliminates important predictors, then the analysis builds sequential models by removing the most important predictors in the sorted list. More specifically, at each model step, the analysis builds a model without the most important predictor from the previous model. The analysis uses that model to recalculate the importance scores for all the predictors that remain in the analysis. The steps of predictor elimination and model construction repeat for the maximum number of elimination steps.

For details on the calculation of the model summary statistics in the Model Selection table, go to Methods and formulas for the model summary in Fit Model and Discover Key Predictors with TreeNet® Regression.