The Relative Variable Importance graph plots the predictors in order of
their effect on model improvement when splits are made on a predictor over the
entire forest. The variable with the highest improvement score is set as the
most important variable, and the other variables follow in order of importance.
Relative variable importance standardizes the importance values for ease of
interpretation. Relative importance is defined as the percent improvement with
respect to the most important predictor, which has an importance of 100%.

Relative importance is calculated by dividing each variable importance
score by the largest importance score of the variables, then we multiply by
100%.

Interpretation

Relative variable importance values range from 0% to 100%. The most
important variable always has a relative importance of 100%. If a variable is
not used in any of the trees, then the variable is not important.

Minitab uses two methods to calculate the relative importance scores for
the variables on the variable importance chart. For
Permutation,
Minitab evaluates how much worse the model performs by validating the model
again with the permuted values of a variable on the chart. For
Gini,
Minitab sums the improvements the variable makes for all of the trees.
Permutation
is the default method for data sets with 5000 or fewer records. Consider
whether to use
Permutation
for larger data sets when the analysis does not take too long and the
identification of important predictors is an important goal.