Select the graphs to display for Fit Model and Discover Key Predictors with TreeNet^® Regression

Select the graphs that you want to display for the analysis.

R-squared vs. number of trees plot

The R-squared vs. number of trees plot shows the relationship between R-squared value and the number of trees in the gradient boosted regression model when the loss function is Squared error or Huber.

Mean absolute deviation vs. number of trees plot

The Mean absolute deviation vs. number of trees plot shows the relationship between mean absolute deviation value and the number of trees when the loss function is Least absolute deviation.

Variable importance chart

The variable importance chart shows the relative importance of the predictors. You can choose whether to display all or some of the important variables. Variables are important when they are used as primary splitters.

• Display all important variables: By default, this chart displays all important variables.
• Display a percentage of important variables: Specify the percentage of important variables to display. Enter a value between 0 and 100.
• Display all predictor variables: Display all predictors whether or not they are important variables.

Fitted vs. actual response value plot

The Fitted vs. actual response value plot shows the fitted Y (response) values versus the actual Y (response) values for both the training and test data sets.

Boxplot of residuals

The Boxplot of residuals shows the residual values or the percent residuals for both the training and test data sets.

One-predictor partial dependence plot for top K important variables, K =

The one predictor partial dependence plots display fits for the top 4 important variables, by default. You can increase or decrease the number of important variables to plot. After you have results, select One-Predictor Plots in the results to show plots for more predictors.

Two-predictor partial dependence plot for top K important variables, K =

The two predictor partial dependence plots display the fits for the top 2 important variables, by default. You can increase or decrease the number of important variables to plot. After you have results, select Two-Predictor Plots in the results to show plots for more pairs of predictors.

For plots with categorical predictors, Minitab plots a scatterplot of the fitted values. For continuous predictors, you can specify Surface, Contour plots, or both.