Skip to main content

Table 5 Analysis by class of a data set in Scenario 1

From: Intervention in prediction measure: a new approach to assessing variable importance for random forests

Measures PVIM IPM
  0 1 G 0 1 G
X 1 4 1 4 3 3 3
X 2 2 3 2 1 1 1
X 3 3 2 3 5 4 5
X 4 1 4 1 2 2 2
X 5 5 5 5 4 5 4
  1. The first column is the name of the variables. The two following columns correspond to the PVIM ranking (CART-RF, mtry = 2) for each class, whereas the third column is the same but calculated globally (labeled as G). The last three columns contain the ranking of the IPM values (CIT-RF, mtry = 5) first by group and the last column computed globally (labeled as G)